Publications
Preprint
Exploration by Exploitation: Curriculum Learning for Reinforcement Learning Agents through Competence-Based Curriculum Policy SearchThe Exploration in AI Today Workshop at ICML 2025@inproceedings{ lee2025exploration, author = { Lee, Tabitha Edith and Ke, Nan Rosemary and Patil, Sarvesh and Dahmani, Annya and Yiu, Eunice and Saleh, Esra’a and Gopnik, Alison and Kroemer, Oliver and Berseth, Glen }, title = { Exploration by Exploitation: Curriculum Learning for Reinforcement Learning Agents through Competence-Based Curriculum Policy Search }, booktitle = { The Exploration in AI Today Workshop at ICML 2025 }, year = { Preprint }, }
Curiosity-Driven Exploration via Temporal Contrastive LearningWorkshop on Reinforcement Learning Beyond Rewards @ Reinforcement Learning Conference 2025@inproceedings{ mohamed2025curiositydriven, author = { Mohamed, Faisal and Ji, Catherine and Eysenbach, Benjamin and Berseth, Glen }, title = { Curiosity-Driven Exploration via Temporal Contrastive Learning }, booktitle = { Workshop on Reinforcement Learning Beyond Rewards @ Reinforcement Learning Conference 2025 }, year = { Preprint }, }
Self-Predictive Representations for Combinatorial Generalization in Behavioral CloningarXiv preprint arXiv:2506.10137@article{ lawson2025self, author = { Lawson, Daniel and Hugessen, Adriana and Cloutier, Charlotte and Berseth, Glen and Khetarpal, Khimya }, title = { Self-Predictive Representations for Combinatorial Generalization in Behavioral Cloning }, journal = { arXiv preprint arXiv:2506.10137 }, year = { Preprint }, }
Is Exploration or Optimization the Problem for Deep Reinforcement Learning?arXiv preprint arXiv:2508.01329@article{ berseth2025exploration, author = { Berseth, Glen }, title = { Is Exploration or Optimization the Problem for Deep Reinforcement Learning? }, journal = { arXiv preprint arXiv:2508.01329 }, year = { Preprint }, }
SegDAC: Segmentation-Driven Actor-Critic for Visual Reinforcement LearningarXiv preprint arXiv:2508.09325@article{ brown2025segdac, author = { Brown, Alexandre and Berseth, Glen }, title = { SegDAC: Segmentation-Driven Actor-Critic for Visual Reinforcement Learning }, journal = { arXiv preprint arXiv:2508.09325 }, year = { Preprint }, }
Inter-Level Cooperation in Hierarchical Reinforcement LearningUnder submission to Journal of Machine Learning Research@article{ CHER, author = { Kreidieh, Abdul Rahman and Berseth, Glen and Parajuli, Brandon Traboco Samyak and Levine, Sergey and Bayen, Alexandre M. }, title = { Inter-Level Cooperation in Hierarchical Reinforcement Learning }, journal = { Under submission to Journal of Machine Learning Research }, year = { Preprint }, }
2025
What Matters for Maximizing Data Reuse In Value-based Deep Reinforcement LearningFinding the Frame Workshop at RLC 2025@inproceedings{ castanyer2025what, author = { Castanyer, Roger Creus and Berseth, Glen and Castro, Pablo Samuel }, title = { What Matters for Maximizing Data Reuse In Value-based Deep Reinforcement Learning }, booktitle = { Finding the Frame Workshop at RLC 2025 }, year = { 2025 }, }
Training PPO-Clip with Parallelized Data Generation: A Case of Fixed-Point ConvergenceInductive Biases in Reinforcement Learning Workshop @ RLC 2025@inproceedings{ honari2025training, author = { Honari, Homayoun and Castanyer, Roger Creus and Castro, Pablo Samuel and Berseth, Glen }, title = { Training PPO-Clip with Parallelized Data Generation: A Case of Fixed-Point Convergence }, booktitle = { Inductive Biases in Reinforcement Learning Workshop @ RLC 2025 }, year = { 2025 }, }
Scalable Tree Search over Graphs with Learned Action Pruning for Power Grid ControlRLC 2025 Workshop on Practical Insights into Reinforcement Learning for Real Systems@inproceedings{ cloutier2025scalable, author = { Cloutier, Florence and Neary, Cyrus and Hugessen, Adriana and Todosijević, Viktor and Kamel, Zina and Berseth, Glen }, title = { Scalable Tree Search over Graphs with Learned Action Pruning for Power Grid Control }, booktitle = { RLC 2025 Workshop on Practical Insights into Reinforcement Learning for Real Systems }, year = { 2025 }, }
Outsourced diffusion sampling: Efficient posterior inference in latent spaces of generative modelsInternational Conference on Machine Learning 2025@article{ venkatraman2025outsourced, author = { Venkatraman, Siddarth and Hasan, Mohsin and Kim, Minsu and Scimeca, Luca and Sendera, Marcin and Bengio, Yoshua and Berseth, Glen and Malkin, Nikolay }, title = { Outsourced diffusion sampling: Efficient posterior inference in latent spaces of generative models }, journal = { International Conference on Machine Learning 2025 }, year = { 2025 }, }
Stable Gradients for Stable Learning at Scale in Deep Reinforcement LearningThe Thirty-ninth Annual Conference on Neural Information Processing Systems (spotlight, top 3%)@inproceedings{ castanyer2025stablegradientsstablelearning, author = { Castanyer, Roger Creus and Obando-Ceron, Johan and Li, Lu and Bacon, Pierre-Luc and Berseth, Glen and Courville, Aaron and Castro, Pablo Samuel }, title = { Stable Gradients for Stable Learning at Scale in Deep Reinforcement Learning }, booktitle = { The Thirty-ninth Annual Conference on Neural Information Processing Systems (spotlight, top 3%) }, year = { 2025 }, }
Mitigating Plasticity Loss in Continual Reinforcement Learning by Reducing ChurnForty-second International Conference on Machine Learning@inproceedings{ tang2025mitigating, author = { Tang, Hongyao and Obando-Ceron, Johan and Castro, Pablo Samuel and Courville, Aaron and Berseth, Glen }, title = { Mitigating Plasticity Loss in Continual Reinforcement Learning by Reducing Churn }, journal = { arXiv preprint arXiv:2506.00592 }, booktitle = { Forty-second International Conference on Machine Learning }, year = { 2025 }, }
RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies9th Annual Conference on Robot Learning@inproceedings{ atreya2025roboarena, author = { Atreya, Pranav and Pertsch, Karl and Lee, Tony and Kim, Moo Jin and Jain, Arhan and Kuramshin, Artur and Eppner, Clemens and Neary, Cyrus and Hu, Edward and Ramos, Fabio and others, }, title = { RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies }, booktitle = { 9th Annual Conference on Robot Learning }, year = { 2025 }, }
Non-Adversarial Inverse Reinforcement Learning via Successor Feature MatchingThe Thirteenth International Conference on Learning Representations@inproceedings{ jain2025nonadversarial, author = { Jain, Arnav Kumar and Wiltzer, Harley and Farebrother, Jesse and Rish, Irina and Berseth, Glen and Choudhury, Sanjiban }, title = { Non-Adversarial Inverse Reinforcement Learning via Successor Feature Matching }, booktitle = { The Thirteenth International Conference on Learning Representations }, year = { 2025 }, }
Enabling Realtime Reinforcement Learning at Scale with Staggered Asynchronous InferenceThe Thirteenth International Conference on Learning Representations@inproceedings{ riemer2024realtime, author = { Riemer, Matthew and Subbaraj, Gopeshh and Berseth, Glen and Rish, Irina }, title = { Enabling Realtime Reinforcement Learning at Scale with Staggered Asynchronous Inference }, booktitle = { The Thirteenth International Conference on Learning Representations }, year = { 2025 }, }
Minimally Invasive Morphology Adaptation via Parameter Efficient Fine-TuningReinforcement Learning Journal@inproceedings{ przystupa2024minimally, author = { Przystupa, Michael and Tang, Hongyao and Phielipp, Mariano and Miret, Santiago and Jägersand, Martin and Berseth, Glen }, title = { Minimally Invasive Morphology Adaptation via Parameter Efficient Fine-Tuning }, booktitle = { Reinforcement Learning Journal }, year = { 2025 }, }
RLeXplore: Accelerating Research in Intrinsically-Motivated Reinforcement LearningTransactions on Machine Learning Research (TMLR)@article{ yuan2024rlexploreacceleratingresearchintrinsicallymotivated, author = { Yuan, Mingqi and Castanyer, Roger Creus and Li, Bo and Jin, Xin and Berseth, Glen and Zeng, Wenjun }, title = { RLeXplore: Accelerating Research in Intrinsically-Motivated Reinforcement Learning }, journal = { Transactions on Machine Learning Research (TMLR) }, year = { 2025 }, }
Adaptive Resolution Residual NetworksTransactions on Machine Learning Research (TMLR)@article{ DemeuleInariant, author = { Demeule, Léa and Sandhu, Mahtab and Berseth, Glen }, title = { Adaptive Resolution Residual Networks }, journal = { Transactions on Machine Learning Research (TMLR) }, year = { 2025 }, }
Towards Improving Exploration through Sibling Augmented GFlowNetsThe Thirteenth International Conference on Learning Representations@inproceedings{ MadanGFNGCRL, author = { Madan, Kanika and Lamb, Alex and Bengio, Emmanuel and Berseth, Glen and Bengio, Yoshua }, title = { Towards Improving Exploration through Sibling Augmented GFlowNets }, booktitle = { The Thirteenth International Conference on Learning Representations }, year = { 2025 }, }
2024
Enhancing Agent Learning through World Dynamics ModelingThe 2024 Conference on Empirical Methods in Natural Language Processing@inproceedings{ sun2024enhancingagentlearningworld, author = { Sun, Zhiyuan and Shi, Haochen and Côté, Marc-Alexandre and Berseth, Glen and Yuan, Xingdi and Liu, Bang }, title = { Enhancing Agent Learning through World Dynamics Modeling }, booktitle = { The 2024 Conference on Empirical Methods in Natural Language Processing }, year = { 2024 }, }
Amortizing intractable inference in diffusion models for vision, language, and controlAdvances in Neural Information Processing Systems@inproceedings{ venkatraman2024amortizing, author = { Venkatraman, Siddarth and Jain, Moksh and Scimeca, Luca and Kim, Minsu and Sendera, Marcin and Hasan, Mohsin and Rowe, Luke and Mittal, Sarthak and Lemos, Pablo and Bengio, Emmanuel and Adam, Alexandre and Rector-Brooks, Jarrid and Bengio, Yoshua and Berseth, Glen and Malkin, Nikolay }, title = { Amortizing intractable inference in diffusion models for vision, language, and control }, booktitle = { Advances in Neural Information Processing Systems }, year = { 2024 }, }
Open X-Embodiment: Robotic Learning Datasets and RT-X Models (ICRA2024 best paper)International Conference on Robotics and Automation (ICRA 2024)@article{ open_x_embodiment_rt_x_2023, author = { Collaboration, Open X-Embodiment and O’Neill, Abby and Rehman, Abdul and Maddukuri, Abhiram and Gupta, Abhishek and Padalkar, Abhishek and Lee, Abraham and Pooley, Acorn and Gupta, Agrim and Mandlekar, Ajay and Jain, Ajinkya and Tung, Albert and Bewley, Alex and Herzog, Alex and Irpan, Alex and Khazatsky, Alexander and Rai, Anant and Gupta, Anchit and Wang, Andrew and Singh, Anikait and Garg, Animesh and Kembhavi, Aniruddha and Xie, Annie and Brohan, Anthony and Raffin, Antonin and Sharma, Archit and Yavary, Arefeh and Jain, Arhan and Balakrishna, Ashwin and Wahid, Ayzaan and Burgess-Limerick, Ben and Kim, Beomjoon and Schölkopf, Bernhard and Wulfe, Blake and Ichter, Brian and Lu, Cewu and Xu, Charles and Le, Charlotte and Finn, Chelsea and Wang, Chen and Xu, Chenfeng and Chi, Cheng and Huang, Chenguang and Chan, Christine and Agia, Christopher and Pan, Chuer and Fu, Chuyuan and Devin, Coline and Xu, Danfei and Morton, Daniel and Driess, Danny and Chen, Daphne and Pathak, Deepak and Shah, Dhruv and Büchler, Dieter and Jayaraman, Dinesh and Kalashnikov, Dmitry and Sadigh, Dorsa and Johns, Edward and Foster, Ethan and Liu, Fangchen and Ceola, Federico and Xia, Fei and Zhao, Feiyu and Stulp, Freek and Zhou, Gaoyue and Sukhatme, Gaurav S. and Salhotra, Gautam and Yan, Ge and Feng, Gilbert and Schiavi, Giulio and Berseth, Glen and Kahn, Gregory and Wang, Guanzhi and Su, Hao and Fang, Hao-Shu and Shi, Haochen and Bao, Henghui and Amor, Heni Ben and Christensen, Henrik I and Furuta, Hiroki and Walke, Homer and Fang, Hongjie and Ha, Huy and Mordatch, Igor and Radosavovic, Ilija and Leal, Isabel and Liang, Jacky and Abou-Chakra, Jad and Kim, Jaehyung and Drake, Jaimyn and Peters, Jan and Schneider, Jan and Hsu, Jasmine and Bohg, Jeannette and Bingham, Jeffrey and Wu, Jeffrey and Gao, Jensen and Hu, Jiaheng and Wu, Jiajun and Wu, Jialin and Sun, Jiankai and Luo, Jianlan and Gu, Jiayuan and Tan, Jie and Oh, Jihoon and Wu, Jimmy and Lu, Jingpei and Yang, Jingyun and Malik, Jitendra and Silvério, João and Hejna, Joey and Booher, Jonathan and Tompson, Jonathan and Yang, Jonathan and Salvador, Jordi and Lim, Joseph J. and Han, Junhyek and Wang, Kaiyuan and Rao, Kanishka and Pertsch, Karl and Hausman, Karol and Go, Keegan and Gopalakrishnan, Keerthana and Goldberg, Ken and Byrne, Kendra and Oslund, Kenneth and Kawaharazuka, Kento and Black, Kevin and Lin, Kevin and Zhang, Kevin and Ehsani, Kiana and Lekkala, Kiran and Ellis, Kirsty and Rana, Krishan and Srinivasan, Krishnan and Fang, Kuan and Singh, Kunal Pratap and Zeng, Kuo-Hao and Hatch, Kyle and Hsu, Kyle and Itti, Laurent and Chen, Lawrence Yunliang and Pinto, Lerrel and Fei-Fei, Li and Tan, Liam and Fan, Linxi "Jim" and Ott, Lionel and Lee, Lisa and Weihs, Luca and Chen, Magnum and Lepert, Marion and Memmel, Marius and Tomizuka, Masayoshi and Itkina, Masha and Castro, Mateo Guaman and Spero, Max and Du, Maximilian and Ahn, Michael and Yip, Michael C. and Zhang, Mingtong and Ding, Mingyu and Heo, Minho and Srirama, Mohan Kumar and Sharma, Mohit and Kim, Moo Jin and Kanazawa, Naoaki and Hansen, Nicklas and Heess, Nicolas and Joshi, Nikhil J and Suenderhauf, Niko and Liu, Ning and Palo, Norman Di and Shafiullah, Nur Muhammad Mahi and Mees, Oier and Kroemer, Oliver and Bastani, Osbert and Sanketi, Pannag R and Miller, Patrick "Tree" and Yin, Patrick and Wohlhart, Paul and Xu, Peng and Fagan, Peter David and Mitrano, Peter and Sermanet, Pierre and Abbeel, Pieter and Sundaresan, Priya and Chen, Qiuyu and Vuong, Quan and Rafailov, Rafael and Tian, Ran and Doshi, Ria and Mart’in-Mart’in, Roberto and Baijal, Rohan and Scalise, Rosario and Hendrix, Rose and Lin, Roy and Qian, Runjia and Zhang, Ruohan and Mendonca, Russell and Shah, Rutav and Hoque, Ryan and Julian, Ryan and Bustamante, Samuel and Kirmani, Sean and Levine, Sergey and Lin, Shan and Moore, Sherry and Bahl, Shikhar and Dass, Shivin and Sonawani, Shubham and Song, Shuran and Xu, Sichun and Haldar, Siddhant and Karamcheti, Siddharth and Adebola, Simeon and Guist, Simon and Nasiriany, Soroush and Schaal, Stefan and Welker, Stefan and Tian, Stephen and Ramamoorthy, Subramanian and Dasari, Sudeep and Belkhale, Suneel and Park, Sungjae and Nair, Suraj and Mirchandani, Suvir and Osa, Takayuki and Gupta, Tanmay and Harada, Tatsuya and Matsushima, Tatsuya and Xiao, Ted and Kollar, Thomas and Yu, Tianhe and Ding, Tianli and Davchev, Todor and Zhao, Tony Z. and Armstrong, Travis and Darrell, Trevor and Chung, Trinity and Jain, Vidhi and Vanhoucke, Vincent and Zhan, Wei and Zhou, Wenxuan and Burgard, Wolfram and Chen, Xi and Wang, Xiaolong and Zhu, Xinghao and Geng, Xinyang and Liu, Xiyuan and Liangwei, Xu and Li, Xuanlin and Lu, Yao and Ma, Yecheng Jason and Kim, Yejin and Chebotar, Yevgen and Zhou, Yifan and Zhu, Yifeng and Wu, Yilin and Xu, Ying and Wang, Yixuan and Bisk, Yonatan and Cho, Yoonyoung and Lee, Youngwoon and Cui, Yuchen and Cao, Yue and Wu, Yueh-Hua and Tang, Yujin and Zhu, Yuke and Zhang, Yunchu and Jiang, Yunfan and Li, Yunshuang and Li, Yunzhu and Iwasawa, Yusuke and Matsuo, Yutaka and Ma, Zehan and Xu, Zhuo and Cui, Zichen Jeff and Zhang, Zichen and Lin, Zipeng }, title = { Open X-Embodiment: Robotic Learning Datasets and RT-X Models (ICRA2024 best paper) }, journal = { International Conference on Robotics and Automation (ICRA 2024) }, year = { 2024 }, }
Improving Deep Reinforcement Learning by Reducing the Chain Effect of Value and Policy ChurnAdvances in Neural Information Processing Systems@inproceedings{ tang2024improving, author = { Tang, Hongyao and Berseth, Glen }, title = { Improving Deep Reinforcement Learning by Reducing the Chain Effect of Value and Policy Churn }, booktitle = { Advances in Neural Information Processing Systems }, year = { 2024 }, }
Simplifying Constraint Inference with Inverse Reinforcement LearningAdvances in Neural Information Processing Systems@inproceedings{ Hugessen2024simpICRL, author = { Hugessen, Adriana and Wiltzer, Harley and Berseth, Glen }, title = { Simplifying Constraint Inference with Inverse Reinforcement Learning }, booktitle = { Advances in Neural Information Processing Systems }, year = { 2024 }, }- DROID: A Large-Scale In-The-Wild Robot Manipulation DatasetRobotics: Science and Systems XX, 2024
@inproceedings{ khazatsky2024droid, author = { Khazatsky, Alexander and Pertsch, Karl and Nair, Suraj and Balakrishna, Ashwin and Dasari, Sudeep and Karamcheti, Siddharth and Nasiriany, Soroush and Srirama, Mohan Kumar and Chen, Lawrence Yunliang and Ellis, Kirsty and Fagan, Peter David and Hejna, Joey and Itkina, Masha and Lepert, Marion and Ma, Yecheng Jason and Miller, Patrick Tree and Wu, Jimmy and Belkhale, Suneel and Dass, Shivin and Ha, Huy and Jain, Arhan and Lee, Abraham and Lee, Youngwoon and Memmel, Marius and Park, Sungjae and Radosavovic, Ilija and Wang, Kaiyuan and Zhan, Albert and Black, Kevin and Chi, Cheng and Hatch, Kyle Beltran and Lin, Shan and Lu, Jingpei and Mercat, Jean and Rehman, Abdul and Sanketi, Pannag R and Sharma, Archit and Simpson, Cody and Vuong, Quan and Walke, Homer Rich and Wulfe, Blake and Xiao, Ted and Yang, Jonathan Heewon and Yavary, Arefeh and Zhao, Tony Z. and Agia, Christopher and Baijal, Rohan and Castro, Mateo Guaman and Chen, Daphne and Chen, Qiuyu and Chung, Trinity and Drake, Jaimyn and Foster, Ethan Paul and Gao, Jensen and Herrera, David Antonio and Heo, Minho and Hsu, Kyle and Hu, Jiaheng and Jackson, Donovon and Le, Charlotte and Li, Yunshuang and Lin, Kevin and Lin, Roy and Ma, Zehan and Maddukuri, Abhiram and Mirchandani, Suvir and Morton, Daniel and Nguyen, Tony and O’Neill, Abigail and Scalise, Rosario and Seale, Derick and Son, Victor and Tian, Stephen and Tran, Emi and Wang, Andrew E. and Wu, Yilin and Xie, Annie and Yang, Jingyun and Yin, Patrick and Zhang, Yunchu and Bastani, Osbert and Berseth, Glen and Bohg, Jeannette and Goldberg, Ken and Gupta, Abhinav and Gupta, Abhishek and Jayaraman, Dinesh and Lim, Joseph J and Malik, Jitendra and Martín-Martín, Roberto and Ramamoorthy, Subramanian and Sadigh, Dorsa and Song, Shuran and Wu, Jiajun and Yip, Michael C. and Zhu, Yuke and Kollar, Thomas and Levine, Sergey and Finn, Chelsea }, title = { DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset }, booktitle = { Robotics: Science and Systems XX, 2024 }, year = { 2024 }, } 
Surprise-Adaptive Intrinsic Motivation for Unsupervised Reinforcement LearningReinforcement Learning Conference 2024@article{ HuggensonSurpriseAdapt, author = { Knatchbull-Hugessen, Adriana and Creus-Castanyer, Roger and Mohamed, Faisal and Berseth, Glen }, title = { Surprise-Adaptive Intrinsic Motivation for Unsupervised Reinforcement Learning }, journal = { Reinforcement Learning Conference 2024 }, year = { 2024 }, }
Reinforcement Learning for Versatile, Dynamic, and Robust Bipedal Locomotion ControlInternational Journal of Robotics Research@article{ li2024reinforcement, author = { Li, Zhongyu and Peng, Xue Bin and Abbeel, Pieter and Levine, Sergey and Berseth, Glen and Sreenath, Koushil }, title = { Reinforcement Learning for Versatile, Dynamic, and Robust Bipedal Locomotion Control }, journal = { International Journal of Robotics Research }, year = { 2024 }, }
Improving Intrinsic Exploration by Creating Stationary ObjectivesThe Twelfth International Conference on Learning Representations@inproceedings{ CreusSOFE2023, author = { Castanyer, Roger Creus and Romoff, Joshua and Berseth, Glen }, title = { Improving Intrinsic Exploration by Creating Stationary Objectives }, booktitle = { The Twelfth International Conference on Learning Representations }, year = { 2024 }, }
Closing the Gap between TD Learning and Supervised Learning - A Generalisation Point of ViewThe Twelfth International Conference on Learning Representations@inproceedings{ Ghugarem2023, author = { Ghugare, Raj and Geist, Matthieu and Eysenbach, Benjamin and Berseth, Glen }, title = { Closing the Gap between TD Learning and Supervised Learning - A Generalisation Point of View }, booktitle = { The Twelfth International Conference on Learning Representations }, year = { 2024 }, }
Intelligent Switching for Reset-Free RLThe Twelfth International Conference on Learning Representations@inproceedings{ PatilTABA, author = { Patil, Darshan and Rajendran, Janarthanan and Berseth, Glen and Chandar, Sarath }, title = { Intelligent Switching for Reset-Free RL }, booktitle = { The Twelfth International Conference on Learning Representations }, year = { 2024 }, }
Searching the space of high-value molecules using reinforcement learning and language modelsThe Twelfth International Conference on Learning Representations@inproceedings{ GhugaremChemRL, author = { Ghugare, Raj and Miret, Santiago and Hugessen, Adriana and Phielipp, Mariano and Berseth, Glen }, title = { Searching the space of high-value molecules using reinforcement learning and language models }, booktitle = { The Twelfth International Conference on Learning Representations }, year = { 2024 }, }
Reasoning with Latent Diffusion in Offline Reinforcement LearningThe Twelfth International Conference on Learning Representations@inproceedings{ VenkatramanDiffuSkill, author = { Venkatraman, Siddarth and Khaitan, Shivesh and Akella, Ravi Tej and Dolan, John and Schneider, Jeff and Berseth, Glen }, title = { Reasoning with Latent Diffusion in Offline Reinforcement Learning }, booktitle = { The Twelfth International Conference on Learning Representations }, year = { 2024 }, }
2023
Maximum State Entropy Exploration using Predecessor and Successor RepresentationsThirty-seventh Conference on Neural Information Processing Systems@inproceedings{ jain2023maximum, author = { Jain, Arnav Kumar and Lehnert, Lucas and Rish, Irina and Berseth, Glen }, title = { Maximum State Entropy Exploration using Predecessor and Successor Representations }, booktitle = { Thirty-seventh Conference on Neural Information Processing Systems }, year = { 2023 }, }
Bootstrapping Adaptive Human-Machine Interfaces with Offline Reinforcement LearningProc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2023)daptive interfaces can help users perform sequential decision-making tasks like robotic teleoperation given noisy, high-dimensional command signals (e.g., from a brain-computer interface). Recent advances in human-in-the-loop machine learning enable such systems to improve by interacting with users, but tend to be limited by the amount of data that they can collect from individual users in practice. In this paper, we propose a reinforcement learning algorithm to address this by training an interface to map raw command signals to actions using a combination of offline pre-training and online fine-tuning. To address the challenges posed by noisy command signals and sparse rewards, we develop a novel method for representing and inferring the user’s long-term intent for a given trajectory. We primarily evaluate our method’s ability to assist users who can only communicate through noisy, high-dimensional input channels through a user study in which 12 participants performed a simulated navigation task by using their eye gaze to modulate a 128-dimensional command signal from their webcam. The results show that our method enables successful goal navigation more often than a baseline directional interface, by learning to denoise user commands signals and provide shared autonomy assistance. We further evaluate on a simulated Sawyer pushing task with eye gaze control, and the Lunar Lander game with simulated user commands, and find that our method improves over baseline interfaces in these domains as well. Extensive ablation experiments with simulated user commands empirically motivate each component of our method.
@inproceedings{ gao2023bootstrapping, author = { Gao, Jensen and Reddy, Siddharth and Berseth, Glen and Dragan, Anca D. and Levine, Sergey }, title = { Bootstrapping Adaptive Human-Machine Interfaces with Offline Reinforcement Learning }, booktitle = { Proc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2023) }, year = { 2023 }, }
Torque-based Deep Reinforcement Learning for Task-and-Robot Agnostic Learning on Bipedal Robots Using Sim-to-Real TransferIEEE Robotics and Automation LettersIn this paper, we review the question of which action space is best suited for controlling a real biped robot in combination with Sim2Real training. Position control has been popular as it has been shown to be more sample efficient and intuitive to combine with other planning algorithms. However, for position control, gain tuning is required to achieve the best possible policy performance. We show that, instead, using a torque-based action space enables task-and-robot agnostic learning with less parameter tuning and mitigates the sim-to-reality gap by taking advantage of torque control’s inherent compliance. Also, we accelerate the torque-based-policy training process by pre-training the policy to remain upright by compensating for gravity. The paper showcases the first successful sim-to-real transfer of a torque-based deep reinforcement learning policy on a real human-sized biped robot.
@article{ kim2023torque, author = { Kim, Donghyeon and Berseth, Glen and Schwartz, Mathew and Park, Jaeheung }, title = { Torque-based Deep Reinforcement Learning for Task-and-Robot Agnostic Learning on Bipedal Robots Using Sim-to-Real Transfer }, journal = { IEEE Robotics and Automation Letters }, year = { 2023 }, }
Robust and Versatile Bipedal Jumping Control through Reinforcement LearningRobotics: Science and Systems XIX, Daegu, Republic of Korea, July 10-14, 2023@inproceedings{ li2023robust, author = { Li, Zhongyu and Peng, Xue Bin and Abbeel, Pieter and Levine, Sergey and Berseth, Glen and Sreenath, Koushil }, title = { Robust and Versatile Bipedal Jumping Control through Reinforcement Learning }, booktitle = { Robotics: Science and Systems XIX, Daegu, Republic of Korea, July 10-14, 2023 }, year = { 2023 }, }
Towards Learning to Imitate from a Single Video DemonstrationJ. Mach. Learn. Res.Agents that can learn to imitate behaviours observed in video–without having direct access to internal state or action information of the observed agent–are more suitable for learning in …
@article{ Berseth2023-tn, author = { Berseth, Glen and Golemo, Florian and Pal, Christopher }, title = { Towards Learning to Imitate from a Single Video Demonstration }, journal = { J. Mach. Learn. Res. }, year = { 2023 }, }
2022
Hierarchical Reinforcement Learning for Precise Soccer Shooting Skills using a Quadrupedal RobotProc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2022)We address the problem of enabling quadrupedal robots to perform precise shooting skills in the real world using reinforcement learning. Developing algorithms to enable a legged robot to shoot a soccer ball to a given target is a challenging problem that combines robot motion control and planning into one task. To solve this problem, we need to consider the dynamics limitation and motion stability during the control of a dynamic legged robot. Moreover, we need to consider motion planning to shoot the hard-to-model deformable ball rolling on the ground with uncertain friction to a desired location. In this paper, we propose a hierarchical framework that leverages deep reinforcement learning to train (a) a robust motion control policy that can track arbitrary motions and (b) a planning policy to decide the desired kicking motion to shoot a soccer ball to a target. We deploy the proposed framework on an A1 quadrupedal robot and enable it to accurately shoot the ball to random targets in the real world.
@inproceedings{ quadsoccer, author = { Ji, Yandong and Li*, Zhongyu and Sun, Yinan and Peng, Xue Bin and Levine, Sergey and Berseth, Glen and Sreenath, Koushil }, title = { Hierarchical Reinforcement Learning for Precise Soccer Shooting Skills using a Quadrupedal Robot }, booktitle = { Proc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2022) }, year = { 2022 }, }
AnyMorph: Learning Transferable Policies By Inferring Agent MorphologyInternation Conference on Machine LearningThe prototypical approach to reinforcement learning involves training policies tailored to a particular agent from scratch for every new morphology. Recent work aims to eliminate the re-training of policies by investigating whether a morphology-agnostic policy, trained on a diverse set of agents with similar task objectives, can be transferred to new agents with unseen morphologies without re-training. This is a challenging problem that required previous approaches to use hand-designed descriptions of the new agent’s morphology. Instead of hand-designing this description, we propose a data-driven method that learns a representation of morphology directly from the reinforcement learning objective. Ours is the first reinforcement learning algorithm that can train a policy to generalize to new agent morphologies without requiring a description of the agent’s morphology in advance. We evaluate our approach on the standard benchmark for agent-agnostic control, and improve over the current state of the art in zero-shot generalization to new agents. Importantly, our method attains good performance without an explicit description of morphology.
@article{ Traboco2022, author = { Trabucco, Brandon and Mariano, Phielipp and Berseth, Glen }, title = { AnyMorph: Learning Transferable Policies By Inferring Agent Morphology }, journal = { Internation Conference on Machine Learning }, year = { 2022 }, }
ASHA: Assistive Teleoperation via Human-in-the-Loop Reinforcement LearningIEEE International Conference on Robotics and Automation (ICRA)@article{ asha, author = { Chen*, Sean and Gao*, Jensen and Reddy, Siddharth and Berseth, Glen and Dragan, Anca D. and Levine, Sergey }, title = { ASHA: Assistive Teleoperation via Human-in-the-Loop Reinforcement Learning }, journal = { IEEE International Conference on Robotics and Automation (ICRA) }, year = { 2022 }, }
CoMPS: Continual Meta Policy SearchInternational Conference on Learning Representations@inproceedings{ berseth2022comps, author = { Berseth, Glen and Zhang, Zhiwei and Zhang, Grace and Finn, Chelsea and Levine, Sergey }, title = { CoMPS: Continual Meta Policy Search }, booktitle = { International Conference on Learning Representations }, year = { 2022 }, }
2021
Explore and Control with Adversarial SurpriseCoRR@article{ fickinger2021explore, author = { Fickinger, Arnaud and Jaques, Natasha and Parajuli, Samyak and Chang, Michael and Rhinehart, Nicholas and Berseth, Glen and Russell, Stuart and Levine, Sergey }, title = { Explore and Control with Adversarial Surprise }, journal = { CoRR }, year = { 2021 }, }
Information is Power: Intrinsic Control via Information CaptureAdvances in Neural Information Processing Systems@inproceedings{ rhinehart2021intrinsic, author = { Rhinehart, Nicholas and Wang, Jenny and Berseth, Glen and Co-Reyes, John D and Hafner, Danijar and Finn, Chelsea and Levine, Sergey }, title = { Information is Power: Intrinsic Control via Information Capture }, booktitle = { Advances in Neural Information Processing Systems }, year = { 2021 }, }
Accelerating Online Reinforcement Learning via Model-Based Meta-LearningLearning to Learn - Workshop at ICLR 2021@inproceedings{ co-reyes2021accelerating, author = { Co-Reyes, John D and Feng, Sarah and Berseth, Glen and Qui, Jie and Levine, Sergey }, title = { Accelerating Online Reinforcement Learning via Model-Based Meta-Learning }, booktitle = { Learning to Learn - Workshop at ICLR 2021 }, year = { 2021 }, }
DisCo RL: Distribution-Conditioned Reinforcement Learning for General-Purpose PoliciesInternational Conference on Robotics and Automation (ICRA 2021)@article{ disco, author = { Nasiriany, Soroush and Pong, Vitchyr H. and Nair, Ashvin and Khazatsky, Alexander and Berseth, Glen and Levine, Sergey }, title = { DisCo RL: Distribution-Conditioned Reinforcement Learning for General-Purpose Policies }, journal = { International Conference on Robotics and Automation (ICRA 2021) }, year = { 2021 }, }
Fully Autonomous Real-World Reinforcement Learning with Applications to Mobile ManipulationConference on Robot Learning (CoRL)@inproceedings{ realmm, author = { Sun, Charles and Devin, Coline and Yang, Brian and Orbik, Jędrzej and Gupta, Abhishek and Berseth, Glen and Levine, Sergey }, title = { Fully Autonomous Real-World Reinforcement Learning with Applications to Mobile Manipulation }, booktitle = { Conference on Robot Learning (CoRL) }, year = { 2021 }, }
SMiRL: Surprise Minimizing Reinforcement Learning in Unstable EnvironmentsInternational Conference on Learning Representations \textbf(Oral, top 1.8% of submissions)@inproceedings{ berseth2021smirl, author = { Berseth, Glen and Geng, Daniel and Devin, Coline Manon and Rhinehart, Nicholas and Finn, Chelsea and Jayaraman, Dinesh and Levine, Sergey }, title = { SMiRL: Surprise Minimizing Reinforcement Learning in Unstable Environments }, booktitle = { International Conference on Learning Representations \textbf(Oral, top 1.8% of submissions) }, year = { 2021 }, }
2020
X2T: Training an X-to-Text Typing Interface with Online Learning from Implicit FeedbackInternational Conference on Learning Representations (ICLR) 2021@article{ x2text, author = { Gao, Jensen and Reddy, Sid and Berseth, Glen and Dragon, Anca and Levine, Sergey }, title = { X2T: Training an X-to-Text Typing Interface with Online Learning from Implicit Feedback }, journal = { International Conference on Learning Representations (ICLR) 2021 }, year = { 2020 }, }
Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal RobotsInternational Conference on Robotics and Automation (ICRA 2021)@article{ robustCassie, author = { Li, Zhongyu and Cheng, Xuxin and Peng, Xue Bin and Abbeel, Pieter and Levine, Sergey and Berseth, Glen and Sreenath, Koushil }, title = { Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots }, journal = { International Conference on Robotics and Automation (ICRA 2021) }, year = { 2020 }, }
Ecological Reinforcement LearningCoRR@article{ coreyes2020ecological, author = { Co-Reyes, John D. and Sanjeev, Suvansh and Berseth, Glen and Gupta, Abhishek and Levine, Sergey }, title = { Ecological Reinforcement Learning }, journal = { CoRR }, year = { 2020 }, }
Gamification of Crowd-Driven Environment DesignIEEE Computer Graphics and Applications@article{ 8964083, author = { Haworth, Michael Brandon and Usman, Muhammad and Schaumann, Davide and Chakraborty, Nilay and Glen, Berseth and Faloutsos, Petros and Kapadia, Mubbasir }, title = { Gamification of Crowd-Driven Environment Design }, journal = { IEEE Computer Graphics and Applications }, year = { 2020 }, }
Morphology-Agnostic Visual Robotic ControlIEEE Robotics and Automation Letters@article{ MAVRC, author = { Yang, Brian and Jayaraman, Dinesh and Glen, Berseth and Efros, Alexei and Levine, Sergey }, title = { Morphology-Agnostic Visual Robotic Control }, journal = { IEEE Robotics and Automation Letters }, year = { 2020 }, }
Deep Integration of Physical Humanoid Control and Crowd Navigation (\textbfbest paper runner up)Proceedings of the 13th International Conference on Motion in Games@inproceedings{ deepcrowds, author = { *, Berseth Glen and Haworth*, Brandon and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Deep Integration of Physical Humanoid Control and Crowd Navigation (\textbfbest paper runner up) }, booktitle = { Proceedings of the 13th International Conference on Motion in Games }, year = { 2020 }, }
2019
Contextual Imagined Goals for Self-Supervised Robotic LearningConference on Robot Learning (CoRL)@inproceedings{ CC-RIG, author = { Nair^*, Ashvin and Bahl^*, Shikhar and Khazatsky^*, Alexander and Pong, Vitchyr and Berseth, Glen and Levine, Sergey }, title = { Contextual Imagined Goals for Self-Supervised Robotic Learning }, booktitle = { Conference on Robot Learning (CoRL) }, year = { 2019 }, }
Interactive Architectural Design with Diverse Solution ExplorationIEEE Transactions on Visualization and Computer Graphics@article{ uDOME, author = { Berseth, Glen and Haworth, Brandon and Usman, Muhammad and Schaumann, Davide and Khayatkhoei, Mahyar and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Interactive Architectural Design with Diverse Solution Exploration }, journal = { IEEE Transactions on Visualization and Computer Graphics }, year = { 2019 }, }
2018
Feedback Control For Cassie With Deep Reinforcement LearningProc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2018)@inproceedings{ cassie, author = { Xie, Zhaoming and Berseth, Glen and Clary, Patrick and Hurst, Jonathan W. and Panne, Michiel }, title = { Feedback Control For Cassie With Deep Reinforcement Learning }, booktitle = { Proc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2018) }, year = { 2018 }, }
Model-Based Action Exploration for Learning Dynamic Motion SkillsProc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2018)@inproceedings{ MBAE, author = { Berseth, Glen and Kyriazis, Alex and Zinin, Ivan and Choi, William and Panne, Michiel }, title = { Model-Based Action Exploration for Learning Dynamic Motion Skills }, booktitle = { Proc. IEEE/RSJ Intl Conf on Intelligent Robots and Systems (IROS 2018) }, year = { 2018 }, }
Progressive Reinforcement Learning with Distillation for Multi-Skilled Motion ControlInternational Conference on Learning Representations (ICLR) 2018@article{ PLAiD, author = { Berseth, Glen and Xie, Cheng and Cernek, Paul and Panne, Michiel Van }, title = { Progressive Reinforcement Learning with Distillation for Multi-Skilled Motion Control }, journal = { International Conference on Learning Representations (ICLR) 2018 }, year = { 2018 }, }
2017
- Evaluating and optimizing evacuation plans for crowd egressIEEE computer graphics and applications
@article{ 8013477, author = { Cassol, Vincius J and Testa, Estêvão Smania and Jung, Cláudio Rosito and Usman, Muhammad and Faloutsos, Petros and Glen, Berseth and Kapadia, Mubbasir and Badler, Norman I and Musse, Soraia Raupp }, title = { Evaluating and optimizing evacuation plans for crowd egress }, journal = { IEEE computer graphics and applications }, year = { 2017 }, } - Perceptual Evaluation of Space in Virtual EnvironmentsProceedings of the Tenth International Conference on Motion in Games
@inproceedings{ Usman:2017:PES:3136457.3136458, author = { Usman, Muhammad and Haworth, Brandon and Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Perceptual Evaluation of Space in Virtual Environments }, booktitle = { Proceedings of the Tenth International Conference on Motion in Games }, year = { 2017 }, } - Crowd Sourced Co-design of Floor Plans Using Simulation Guided GamesProceedings of the Tenth International Conference on Motion in Games
@inproceedings{ Chakraborty:2017:CSC:3136457.3136463, author = { Chakraborty, Nilay and Haworth, Brandon and Usman, Muhammad and Glen, Berseth and Faloutsos, Petros and Kapadia, Mubbasir }, title = { Crowd Sourced Co-design of Floor Plans Using Simulation Guided Games }, booktitle = { Proceedings of the Tenth International Conference on Motion in Games }, year = { 2017 }, } - On density flow relationships during crowd evacuationComputer Animation and Virtual Worlds 28 (3-4)
@article{ density_flow_relationship, author = { Haworth, Brandon and Usman, Muhammad and Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { On density flow relationships during crowd evacuation }, journal = { Computer Animation and Virtual Worlds 28 (3-4) }, year = { 2017 }, } 
DeepLoco: Dynamic Locomotion Skills Using Hierarchical Deep Reinforcement LearningSIGGRAPH 2017@article{ peng2017hikeRL, author = { Peng, Xue Bin and Glen, Berseth and Yin, KangKang and Panne, Michiel }, title = { DeepLoco: Dynamic Locomotion Skills Using Hierarchical Deep Reinforcement Learning }, journal = { SIGGRAPH 2017 }, year = { 2017 }, }- Understanding spatial perception and visual modes in the review of architectural designsProceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation, 31
@inproceedings{ usmanGI2017, author = { Usman, Muhammad and Haworth, Brandon and Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Understanding spatial perception and visual modes in the review of architectural designs }, booktitle = { Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation, 31 }, year = { 2017 }, } - CODE: Crowd Optimized Design of EnvironmentsComputer Animation and Virtual Worlds
@article{ CODE2017, author = { Haworth, Brandon and Usman, Muhammad and Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { CODE: Crowd Optimized Design of Environments }, journal = { Computer Animation and Virtual Worlds }, year = { 2017 }, }
2016
- ACCLMesh: curvature-based navigation mesh generationComputer Animation and Virtual Worlds
@article{ berseth2016acclmesh, author = { Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { ACCLMesh: curvature-based navigation mesh generation }, journal = { Computer Animation and Virtual Worlds }, year = { 2016 }, } 
Terrain-adaptive locomotion skills using deep reinforcement learningACM Transactions on Graphics (TOG)@article{ peng2016terrain, author = { Peng, Xue Bin and Glen, Berseth and Panne, Michiel }, title = { Terrain-adaptive locomotion skills using deep reinforcement learning }, journal = { ACM Transactions on Graphics (TOG) }, year = { 2016 }, }- Using synthetic crowds to inform building pillar placementsVirtual Humans and Crowds for Immersive Environments (VHCIE), IEEE
@inproceedings{ haworth2016using, author = { Haworth, Brandon and Usman, Muhammad and Glen, Berseth and Khayatkhoei, Mahyar and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Using synthetic crowds to inform building pillar placements }, booktitle = { Virtual Humans and Crowds for Immersive Environments (VHCIE), IEEE }, year = { 2016 }, } - Towards Computer Assisted Crowd Aware Architectural DesignCHI ’16 Extended Abstracts
@inproceedings{ codeLBW, author = { Haworth, Brandon and Usman, Muhammad and Glen, Berseth and Khayatkhoei, Mahyar and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Towards Computer Assisted Crowd Aware Architectural Design }, booktitle = { CHI ’16 Extended Abstracts }, year = { 2016 }, }
2015
Dynamic Terrain Traversal Skills Using Reinforcement LearningACM Trans. Graph.@article{ Peng:2015:DTT:2809654.2766910, author = { Peng, Xue Bin and Glen, Berseth and Panne, Michiel }, title = { Dynamic Terrain Traversal Skills Using Reinforcement Learning }, journal = { ACM Trans. Graph. }, year = { 2015 }, }- Environment optimization for crowd evacuationComputer Animation and Virtual Worlds
The layout of a building, real or virtual, affects the flow patterns of its intended users. It is well established, for example, that the placement of pillars at proper locations can often facilitate pedestrian flow during the evacuation of a building. Such considerations are therefore important for architects, game level developers, and others whose domains involve agents navigating through buildings. In this paper, we take the first steps towards developing a simulation framework that can be used to study the optimal placement of architectural elements, such as pillars or doors, for the purposes of facilitating dense pedestrian flow during the evacuation of a building. In particular, we show that the steering algorithms used to model the local navigation abilities of the agents significantly affect the results, which motivates the need for a statistically valid approach and further study. Copyright © 2015 John Wiley & Sons, Ltd.
@article{ CAV:CAV1652, author = { Glen, Berseth and Usman, Muhammad and Haworth, Brandon and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Environment optimization for crowd evacuation }, journal = { Computer Animation and Virtual Worlds }, year = { 2015 }, } - Robust Space-Time Footsteps for Agent-Based Steering (best short paper)Computer Animation and Virtual Worlds
Recent agent-based steering methods abandon the standard particle abstraction of an agent’s locomotion abilities and employ more complex models from timed footsteps to physics-based controllers. These models often provide the action space of an optimal search method that plans a sequence of steering actions for each agent that minimize a performance criterion. The transition from particle-based models to more complex models is not straightforward and gives rise to a number of technical challenges. For example, a disk geometry is constant, symmetric and convex, while a footstep model maybe non-convex and dynamic. In this paper, we identify general challenges associated with footstep-based steering approaches and present a new space-time footstep planning steering model that is robust to challenging scenario configurations.
@article{ CASA2015:RobustFootsteps, author = { Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Robust Space-Time Footsteps for Agent-Based Steering (best short paper) }, journal = { Computer Animation and Virtual Worlds }, year = { 2015 }, } - ACCLMesh: Curvature-based Navigation Mesh GenerationProceedings of the 8th ACM SIGGRAPH Conference on Motion in Games
@inproceedings{ Berseth:2015:ACN:2822013.2822043, author = { Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { ACCLMesh: Curvature-based Navigation Mesh Generation }, booktitle = { Proceedings of the 8th ACM SIGGRAPH Conference on Motion in Games }, year = { 2015 }, } - Evaluating and Optimizing Level of Service for Crowd EvacuationsProceedings of the 8th ACM SIGGRAPH Conference on Motion in Games
@inproceedings{ Haworth:2015:EOL:2822013.2822040, author = { Haworth, Brandon and Usman, Muhammad and Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Evaluating and Optimizing Level of Service for Crowd Evacuations }, booktitle = { Proceedings of the 8th ACM SIGGRAPH Conference on Motion in Games }, year = { 2015 }, }
2014
- Characterizing and Optimizing Game Level DifficultyProceedings of the Seventh International Conference on Motion in Games
@inproceedings{ Berseth:2014:COG:2668064.2668100, author = { Glen, Berseth and Haworth, M. Brandon and Kapadia, Mubbasir and Faloutsos, Petros }, title = { Characterizing and Optimizing Game Level Difficulty }, booktitle = { Proceedings of the Seventh International Conference on Motion in Games }, year = { 2014 }, } - SteerFit: Automated Parameter Fitting for Steering AlgorithmsEurographics/ ACM SIGGRAPH Symposium on Computer Animation
In the context of crowd simulation, there is a diverse set of algorithms that model steering. The performance of steering approaches, both in terms of quality of results and computational efficiency, depends on internal parameters that are manually tuned to satisfy application-specific requirements. This paper investigates the effect that these parameters have on an algorithm’s performance. Using three representative steering algorithms and a set of established performance criteria, we perform a number of large scale optimization experiments that optimize an algorithm’s parameters for a range of objectives. For example, our method automatically finds optimal parameters to minimize turbulence at bottlenecks, reduce building evacuation times, produce emergent patterns, and increase the computational efficiency of an algorithm. We also propose using the Pareto Optimal front as an efficient way of modelling optimal relationships between multiple objectives, and demonstrate its effectiveness by estimating optimal parameters for interactively defined combinations of the associated objectives. The proposed methodologies are general and can be applied to any steering algorithm using any set of performance criteria.
@inproceedings{ SCA14:113-122:2014, author = { Berseth, Glen and Kapadia, Mubbasir and Haworth, Brandon and Faloutsos, Petros }, title = { SteerFit: Automated Parameter Fitting for Steering Algorithms }, journal = { Eurographics/ ACM SIGGRAPH Symposium on Computer Animation }, booktitle = { Eurographics/ ACM SIGGRAPH Symposium on Computer Animation }, year = { 2014 }, }
2013
- SteerPlex: Estimating Scenario Complexity for Simulated CrowdsProceedings of Motion on Games
@inproceedings{ Berseth:2013:SES:2522628.2522650, author = { Glen, Berseth and Kapadia, Mubbasir and Faloutsos, Petros }, title = { SteerPlex: Estimating Scenario Complexity for Simulated Crowds }, booktitle = { Proceedings of Motion on Games }, year = { 2013 }, }