skip to main content


Title: Towards multi-agent autonomous racing with the Deepracing framework.
Multi-agent autonomous racing is a challenging problem for autonomous vehicles due to the split-second, and complex decisions that vehicles must continuously make during a race. The presence of other agents on the track requires continuous monitoring of the ego vehicle’s surroundings, and necessitates predicting the behavior of other vehicles so the ego can quickly react to a changing environment with informed decisions. In our previous work we have developed the DeepRacing AI framework for autonomous formula one racing. Our DeepRacing framework was the first implementation to use the highly photorealisitc Formula One game as a simulation testbed for autonomous racing. We have successfully demonstrated single agent high speed autonomous racing using Bezier curve trajectories. In this paper, we extend the capabilities of the DeepRacing framework towards multi-agent autonomous racing. To do so, we first develop and learn a virtual camera model from game data that the user can configure to emulate the presence of a camera sensor on the vehicle. Next we propose and train a deep recurrent neural network that can predict the future poses of opponent agents in the field of view of the virtual camera using vehicles position, velocity, and heading data with respect to the ego vehicle racecar. We demonstrate early promising results for both these contributions in the game. These added features will extend the DeepRacing framework to become more suitable for multi-agent autonomous racing algorithm development  more » « less
Award ID(s):
2046582
NSF-PAR ID:
10320142
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
International Conference on Robotics and Automation (ICRA) - Workshop on Opportunities and Challenges with Autonomous Racing
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Connected Autonomous Vehicles (CAVs) are expected to enable reliable, efficient, and intelligent transportation systems. Most motion planning algorithms for multi-agent systems implicitly assume that all vehicles/agents will execute the expected plan with a small error and evaluate their safety constraints based on this fact. This assumption, however, is hard to keep for CAVs since they may have to change their plan (e.g., to yield to another vehicle) or are forced to stop (e.g., A CAV may break down). While it is desired that a CAV never gets involved in an accident, it may be hit by other vehicles and sometimes, preventing the accident is impossible (e.g., getting hit from behind while waiting behind the red light). Responsibility-Sensitive Safety (RSS) is a set of safety rules that defines the objective of CAV to blame, instead of safety. Thus, instead of developing a CAV algorithm that will avoid any accident, it ensures that the ego vehicle will not be blamed for any accident it is a part of. Original RSS rules, however, are hard to evaluate for merge, intersection, and unstructured road scenarios, plus RSS rules do not prevent deadlock situations among vehicles. In this paper, we propose a new formulation for RSS rules that can be applied to any driving scenario. We integrate the proposed RSS rules with the CAV’s motion planning algorithm to enable cooperative driving of CAVs. We use Control Barrier Functions to enforce safety constraints and compute the energy optimal trajectory for the ego CAV. Finally, to ensure liveness, our approach detects and resolves deadlocks in a decentralized manner. We have conducted different experiments to verify that the ego CAV does not cause an accident no matter when other CAVs slow down or stop. We also showcase our deadlock detection and resolution mechanism using our simulator. Finally, we compare the average velocity and fuel consumption of vehicles when they drive autonomously with the case that they are autonomous and connected. 
    more » « less
  2. Self-driving autonomous vehicles (AVs) have recently gained popularity as a research topic. The safety of AVs is exceptionally important as failure in the design of an AV could lead to catastrophic consequences. AV systems are highly heterogeneous with many different and complex components, so it is difficult to perform end-to-end testing. One solution to this dilemma is to evaluate AVs using simulated racing competition. In this thesis, we present a simulated autonomous racing competition, Generalized RAcing Intelligence Competition (GRAIC). To compete in GRAIC, participants need to submit their controller files which are deployed on a racing ego-vehicle on different race tracks. To evaluate the submitted controller, we also developed a testing pipeline, Autonomous System Operations (AutOps). AutOps is an automated, scalable, and fair testing pipeline developed using software engineering techniques such as continuous integration, containerization, and serverless computing. In order to evaluate the submitted controller in non-trivial circumstances, we populate the race tracks with scenarios, which are pre-defined traffic situations commonly seen in the real road. We present a dynamic scenario testing strategy that generates new scenarios based on results of the ego-vehicle passing through previous scenarios. 
    more » « less
  3. null (Ed.)
    Balancing performance and safety is crucial to deploying autonomous vehicles in multi-agent environments. In particular, autonomous racing is a domain that penalizes safe but conservative policies, highlighting the need for robust, adaptive strategies. Current approaches either make simplifying assumptions about other agents or lack robust mechanisms for online adaptation. This work makes algorithmic contributions to both challenges. First, to generate a realistic, diverse set of opponents, we develop a novel method for self-play based on replica-exchange Markov chain Monte Carlo. Second, we propose a distributionally robust bandit optimization procedure that adaptively adjusts risk aversion relative to uncertainty in beliefs about opponents’ behaviors. We rigorously quantify the tradeoffs in performance and robustness when approximating these computations in real-time motion-planning, and we demonstrate our methods experimentally on autonomous vehicles that achieve scaled speeds comparable to Formula One racecars. 
    more » « less
  4. With the development of sensing and communica- tion technologies in networked cyber-physical systems (CPSs), multi-agent reinforcement learning (MARL)-based methodolo- gies are integrated into the control process of physical systems and demonstrate prominent performance in a wide array of CPS domains, such as connected autonomous vehicles (CAVs). However, it remains challenging to mathematically characterize the improvement of the performance of CAVs with commu- nication and cooperation capability. When each individual autonomous vehicle is originally self-interest, we can not assume that all agents would cooperate naturally during the training process. In this work, we propose to reallocate the system’s total reward efficiently to motivate stable cooperation among autonomous vehicles. We formally define and quantify how to reallocate the system’s total reward to each agent under the proposed transferable utility game, such that communication- based cooperation among multi-agents increases the system’s total reward. We prove that Shapley value-based reward reallocation of MARL locates in the core if the transferable utility game is a convex game. Hence, the cooperation is stable and efficient and the agents should stay in the coalition or the cooperating group. We then propose a cooperative policy learning algorithm with Shapley value reward reallocation. In experiments, compared with several literature algorithms, we show the improvement of the mean episode system reward of CAV systems using our proposed algorithm. 
    more » « less
  5. Abstract

    The imminent deployment of autonomous vehicles requires algorithms capable of making moral decisions in relevant traffic situations. Some scholars in the ethics of autonomous vehicles hope to align such intelligent systems with human moral judgment. For this purpose, studies like the Moral Machine Experiment have collected data about human decision-making in trolley-like traffic dilemmas. This paper first argues that the trolley dilemma is an inadequate experimental paradigm for investigating traffic moral judgments because it does not include agents’ character-based considerations and is incapable of facilitating the investigation of low-stakes mundane traffic scenarios. In light of the limitations of the trolley paradigm, this paper presents an alternative experimental framework that addresses these issues. The proposed solution combines the creation of mundane traffic moral scenarios using virtual reality and the Agent-Deed-Consequences (ADC) model of moral judgment as a moral-psychological framework. This paradigm shift potentially increases the ecological validity of future studies by providing more realism and incorporating character considerations into traffic actions.

     
    more » « less