{"663534":{"#nid":"663534","#data":{"type":"event","title":"MS Proposal by Joshua Kuperman","body":[{"value":"\u003Cp\u003EJoshua Kuperman\u003Cbr \/\u003E\r\n[Advisor: Dr. Evangelos Theodorou]\u003Cbr \/\u003E\r\nwill propose a master\u0026rsquo;s thesis entitled,\u003Cbr \/\u003E\r\nIntegrating Perception into Safe Differentiable Control\u003Cbr \/\u003E\r\nOn\u003Cbr \/\u003E\r\nTuesday, December 13 at 10:00 a.m.\u003Cbr \/\u003E\r\nWeber SS\u0026amp;T 200\u003Cbr \/\u003E\r\nAbstract\u003Cbr \/\u003E\r\nA great challenge exists at the intersection of perception and controls \u0026ndash; integrating the uncertainty\u003Cbr \/\u003E\r\npresent in perception-based state and obstacle estimation into safe control and trajectory optimization.\u003Cbr \/\u003E\r\nThis thesis proposes a model-based learning framework with a policy defined by a safe differentiable\u003Cbr \/\u003E\r\noptimal controller. We will leverage many of the ideas of the world model, an unsupervised\u003Cbr \/\u003E\r\nreinforcement learning technique that has achieved human-level or better-than-human performance on\u003Cbr \/\u003E\r\nmany Atari games. Specifically, we intend on training a variational autoencoder, a common\u003Cbr \/\u003E\r\nunsupervised image processing technique, to learn a latent space representation that is decoded into a\u003Cbr \/\u003E\r\nform that a safety function can be defined on, such as a depth map or occupancy grid. We will learn the\u003Cbr \/\u003E\r\ndynamics of this latent space, as well as a mapping from the latent space directly to a safety function, to\u003Cbr \/\u003E\r\nprovide a differentiable controller with information on how the agent and the environment changes\u003Cbr \/\u003E\r\nover time as a function of the control actions. The controller will have the safety function embedded\u003Cbr \/\u003E\r\ninto the dynamics using barrier states. The barrier state (BaS), and its discrete counterpart (DBaS), is a\u003Cbr \/\u003E\r\nrecently developed method of embedding the safety of a system into the dynamics, providing greater\u003Cbr \/\u003E\r\nsafety information than penalty methods, a regularizing effect, and safety guarantees to complex\u003Cbr \/\u003E\r\ndynamical systems in environments with many obstacles. Tolerant discrete barrier states (TDBaS)\u003Cbr \/\u003E\r\napproximate the safety guarantees of DBaS while improving exploration, allowing for unsafe initial state\u003Cbr \/\u003E\r\ntrajectories, and providing several parameters that can be intuitively tuned for any application. This\u003Cbr \/\u003E\r\nthesis explores how differentiable trajectory optimization can learn these TDBaS safety parameters\u003Cbr \/\u003E\r\ngiven safety uncertainty in a reinforcement learning setting with limited supervision. Towards this end,\u003Cbr \/\u003E\r\nwe will explore a variety of strategies and structures for the encoder-decoder network, the dynamics\u003Cbr \/\u003E\r\nnetwork, the safety function network, and the differentiable controller such as Parametric Differentiable\u003Cbr \/\u003E\r\nDynamic Programming (PDDP), Pontryagin Differentiable Programming (PDP), Barrier Nets, and\u003Cbr \/\u003E\r\nDifferentiable MPC. We will test this framework in simulation, and if time allows, on hardware in the\u003Cbr \/\u003E\r\nIndoor Flight Laboratory or Robotarium.\u003Cbr \/\u003E\r\nCommittee\u003Cbr \/\u003E\r\n\u0026bull; Prof. Evangelos Theodorou \u0026ndash; School of Aerospace Engineering (advisor)\u003Cbr \/\u003E\r\n\u0026bull; Prof. Kyriakos G. Vamvoudakis \u0026ndash; School of Aerospace Engineering\u003Cbr \/\u003E\r\n\u0026bull; Prof. Patricio Vela \u0026ndash; School of Electrical and Computer Engineering\u003C\/p\u003E\r\n","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"Integrating Perception into Safe Differentiable Control"}],"uid":"27707","created_gmt":"2022-11-30 16:33:46","changed_gmt":"2022-11-30 16:33:46","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2022-12-13T10:00:00-05:00","event_time_end":"2022-12-13T12:00:00-05:00","event_time_end_last":"2022-12-13T12:00:00-05:00","gmt_time_start":"2022-12-13 15:00:00","gmt_time_end":"2022-12-13 17:00:00","gmt_time_end_last":"2022-12-13 17:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"166866","name":"MS Proposal"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78761","name":"Faculty\/Staff"},{"id":"78771","name":"Public"},{"id":"78751","name":"Undergraduate students"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}