{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/integrating-kinematics-and-environment","title":"Integrating kinematics and environment context into deep inverse reinforcement learning for predicting off-road vehicle trajectories","arxiv_id":"1810.07225","date":"2018-10-16","proceeding":null,"authors":["Yanfu Zhang","Wenshan Wang","Rogerio Bonatti","Daniel Maturana","Sebastian Scherer"],"abstract":"Predicting the motion of a mobile agent from a third-person perspective is an\nimportant component for many robotics applications, such as autonomous\nnavigation and tracking. With accurate motion prediction of other agents,\nrobots can plan for more intelligent behaviors to achieve specified objectives,\ninstead of acting in a purely reactive way. Previous work addresses motion\nprediction by either only filtering kinematics, or using hand-designed and\nlearned representations of the environment. Instead of separating kinematic and\nenvironmental context, we propose a novel approach to integrate both into an\ninverse reinforcement learning (IRL) framework for trajectory prediction.\nInstead of exponentially increasing the state-space complexity with kinematics,\nwe propose a two-stage neural network architecture that considers motion and\nenvironment together to recover the reward function. The first-stage network\nlearns feature representations of the environment using low-level LiDAR\nstatistics and the second-stage network combines those learned features with\nkinematics data. We collected over 30 km of off-road driving data and validated\nexperimentally that our method can effectively extract useful environmental and\nkinematic features. We generate accurate predictions of the distribution of\nfuture trajectories of the vehicle, encoding complex behaviors such as\nmulti-modal distributions at road intersections, and even show different\npredictions at the same intersection depending on the vehicle's speed.","url_abs":"http://arxiv.org/abs/1810.07225v1","url_pdf":"http://arxiv.org/pdf/1810.07225v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"integrating-kinematics-and-environment","repo_url":"https://github.com/yfzhang/vehicle-motion-forecasting","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"autonomous-navigation","task_name":"Autonomous Navigation"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"trajectory-prediction","task_name":"Trajectory Prediction"},{"task_slug":"motion-prediction","task_name":"motion prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.07225","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}