Urban Driving with Conditional Imitation Learning
Authors:
Jeffrey Hawke,
Richard Shen,
Corina Gurau,
Siddharth Sharma,
Daniele Reda,
Nikolay Nikolov,
Przemyslaw Mazur,
Sean Micklethwaite,
Nicolas Griffiths,
Amar Shah,
Alex Kendall
Abstract:
Hand-crafting generalised decision-making rules for real-world urban autonomous driving is hard. Alternatively, learning behaviour from easy-to-collect human driving demonstrations is appealing. Prior work has studied imitation learning (IL) for autonomous driving with a number of limitations. Examples include only performing lane-following rather than following a user-defined route, only using a…
▽ More
Hand-crafting generalised decision-making rules for real-world urban autonomous driving is hard. Alternatively, learning behaviour from easy-to-collect human driving demonstrations is appealing. Prior work has studied imitation learning (IL) for autonomous driving with a number of limitations. Examples include only performing lane-following rather than following a user-defined route, only using a single camera view or heavily cropped frames lacking state observability, only lateral (steering) control, but not longitudinal (speed) control and a lack of interaction with traffic. Importantly, the majority of such systems have been primarily evaluated in simulation - a simple domain, which lacks real-world complexities. Motivated by these challenges, we focus on learning representations of semantics, geometry and motion with computer vision for IL from human driving demonstrations. As our main contribution, we present an end-to-end conditional imitation learning approach, combining both lateral and longitudinal control on a real vehicle for following urban routes with simple traffic. We address inherent dataset bias by data balancing, training our final policy on approximately 30 hours of demonstrations gathered over six months. We evaluate our method on an autonomous vehicle by driving 35km of novel routes in European urban streets.
△ Less
Submitted 5 December, 2019; v1 submitted 30 November, 2019;
originally announced December 2019.
Coincident Molecular Auxeticity and Negative Order Parameter in a Liquid Crystal Elastomer
Authors:
Devesh Mistry,
Simon D. Connell,
Stuart L. Micklethwaite,
Philip B. Morgan,
John H. Clamp,
Helen F. Gleeson
Abstract:
"Auxetic" materials have the counter-intuitive property of expanding rather than contracting perpendicular to an applied stretch, formally they have negative Poisson's Ratios (PRs).[1,2] This results in properties such as enhanced energy absorption and indentation resistance, which means that auxetics have potential for applications in areas from aerospace to biomedical industries.[3,4] Existing s…
▽ More
"Auxetic" materials have the counter-intuitive property of expanding rather than contracting perpendicular to an applied stretch, formally they have negative Poisson's Ratios (PRs).[1,2] This results in properties such as enhanced energy absorption and indentation resistance, which means that auxetics have potential for applications in areas from aerospace to biomedical industries.[3,4] Existing synthetic auxetics are all created by carefully structuring porous geometries from positive PR materials. Crucially, their geometry causes the auxeticity.[3,4] The necessary porosity weakens the material compared to the bulk and the structure must be engineered, for example, by using resource-intensive additive manufacturing processes.[1,5] A longstanding goal for researchers has been the development of a synthetic material that has intrinsic auxetic behaviour. Such "molecular auxetics" would avoid porosity-weakening and their very existence implies chemical tuneability.[1,4-9] However molecular auxeticity has never previously been proven for a synthetic material.[6,7] Here we present a synthetic molecular auxetic based on a monodomain liquid crystal elastomer (LCE). When stressed perpendicular to the alignment direction, the LCE becomes auxetic at strains greater than approximately 0.8 with a minimum PR of -0.8. The critical strain for auxeticity coincides with the occurrence of a negative liquid crystal order parameter (LCOP). We show the auxeticity agrees with theoretical predictions derived from the Warner and Terentjev theory of LCEs.[10] This demonstration of a synthetic molecular auxetic represents the origin of a new approach to producing molecular auxetics with a range of physical properties and functional behaviours. Further, it demonstrates a novel feature of LCEs and a route for realisation of the molecular auxetic technologies that have been proposed over the years.
△ Less
Submitted 11 June, 2018;
originally announced July 2018.
