Latent Imagination Facilitates Zero-Shot Transfer in Autonomous Racing

Brunnbauer, Axel; Luigi, Berducci,; Andreas, Brandstätter,; Lechner, Mathias; Hasani, Ramin; Rus, Daniela; Grosu, Radu

doi:10.48550/arxiv.2103.04909

Cited by 2 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Simulation has emerged as a viable candidate to overcome this challenge and render a continuum of scenarios for learning in the presence of other objects and agents in the environment. Works that learn object avoidance in simulation have leveraged both imitation learning [54] as well as reinforcement learning [8,28,11,58,19] but often face limited to no deployment capabilities in reality due to large sim-to-real gaps present in model-based simulation. In this work, we leverage recent advances in data-driven simulation [3,30,46,4] to overcome the sim-to-real gap to learn robust end-to-end controllers capable of transferring to real scenarios with other agents.…”

Section: Related Workmentioning

confidence: 99%

Differentiable Control Barrier Functions for Vision-based End-to-End Autonomous Driving

Xiao¹,

Wang²,

Makram³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Guaranteeing safety of perception-based learning systems is challenging due to the absence of ground-truth state information unlike in state-aware control scenarios. In this paper, we introduce a safety guaranteed learning framework for visionbased end-to-end autonomous driving. To this end, we design a learning system equipped with differentiable control barrier functions (dCBFs) that is trained end-to-end by gradient descent. Our models are composed of conventional neural network architectures and dCBFs. They are interpretable at scale, achieve great test performance under limited training data, and are safety guaranteed in a series of autonomous driving scenarios such as lane keeping and obstacle avoidance. We evaluated our framework in a sim-to-real environment, and tested on a real autonomous car, achieving safe lane following and obstacle avoidance via Augmented Reality (AR) and real parked vehicles.

show abstract

Section: Related Workmentioning

confidence: 99%

Differentiable Control Barrier Functions for Vision-based End-to-End Autonomous Driving

Xiao¹,

Wang²,

Makram³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…In the context of control from pixel images, world models [25] proved that it is possible to learn accurate dynamic models for POMDPs by using noisy high-dimensional observations instead of accurate states. Their application in planning [27], and later in policy learning [26], have achieved the new state-of-the-art performance in many benchmarks and were recently applied to real-world robots [12].…”

Section: Related Workmentioning

confidence: 99%

Safe Policy Improvement in Constrained Markov Decision Processes

Luigi

Grosu

2022

Leveraging Applications of Formal Methods, Verification and Validation. Verification Principles

View full text Add to dashboard Cite

The automatic synthesis of a policy through reinforcement learning (RL) from a given set of formal requirements depends on the construction of a reward signal and consists of the iterative application of many policy-improvement steps. The synthesis algorithm has to balance target, safety, and comfort requirements in a single objective and to guarantee that the policy improvement does not increase the number of safety-requirements violations, especially for safety-critical applications. In this work, we present a solution to the synthesis problem by solving its two main challenges: reward-shaping from a set of formal requirements and safe policy update. For the first, we propose an automatic rewardshaping procedure, defining a scalar reward signal compliant with the task specification. For the second, we introduce an algorithm ensuring that the policy is improved in a safe fashion, with high-confidence guarantees. We also discuss the adoption of a model-based RL algorithm to efficiently use the collected data and train a model-free agent on the predicted trajectories, where the safety violation does not have the same impact as in the real world. Finally, we demonstrate in standard control benchmarks that the resulting learning procedure is effective and robust even under heavy perturbations of the hyperparameters.

show abstract

Latent Imagination Facilitates Zero-Shot Transfer in Autonomous Racing

Cited by 2 publications

References 0 publications

Differentiable Control Barrier Functions for Vision-based End-to-End Autonomous Driving

Differentiable Control Barrier Functions for Vision-based End-to-End Autonomous Driving

Safe Policy Improvement in Constrained Markov Decision Processes

Contact Info

Product

Resources

About