Dynamic Environment Prediction in Urban Scenes using Recurrent Representation Learning

Itkina, Masha; Driggs-Campbell, Katherine; Kochenderfer, Mykel J.

doi:10.1109/itsc.2019.8917271

Cited by 32 publications

(64 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another approach to address the problem of highdimensional state space is to use a mid-level representation, such as a bird's eye view of the current scene, as the state space. The mid-level representations can be constructed by engineered perception modules or by learning the mapping from sensor measurements to the mid-level representation [11]. Such representations are useful because they can capture the entire traffic environment around the vehicle in an interpretable fashion, but their dimensionality will still be high.…”

Section: Related Workmentioning

confidence: 99%

Increasing the Efficiency of Policy Learning for Autonomous Vehicles by Multi-Task Representation Learning

Kargar

Kyrki

2022

IEEE Trans. Intell. Veh.

View full text Add to dashboard Cite

Driving in a dynamic, multi-agent, and complex urban environment is a difficult task requiring a complex decision-making policy. The learning of such a policy requires a state representation that can encode the entire environment. Mid-level representations that encode a vehicle's environment as images have become a popular choice. Still, they are quite highdimensional, limiting their use in data-hungry approaches such as reinforcement learning. In this article, we propose to learn a lowdimensional and rich latent representation of the environment by leveraging the knowledge of relevant semantic factors. To do this, we train an encoder-decoder deep neural network to predict multiple application-relevant factors such as the trajectories of other agents and the ego car. Furthermore, we propose a hazard signal based on other vehicles' future trajectories and the planned route which is used in conjunction with the learned latent representation as input to a down-stream policy. We demonstrate that using the multi-head encoder-decoder neural network results in a more informative representation than a standard single-head model. In particular, the proposed representation learning and the hazard signal help reinforcement learning to learn faster, with increased performance and less data than baseline methods.

show abstract

Section: Related Workmentioning

confidence: 99%

Increasing the Efficiency of Policy Learning for Autonomous Vehicles by Multi-Task Representation Learning

Kargar

Kyrki

2022

IEEE Trans. Intell. Veh.

View full text Add to dashboard Cite

show abstract

“…For pedestrians, also linear quadratic regulator-based models are used [58]. Probability distributions can be represented as occupancy grids, which are obtained through machine learning [59]- [62] or Markov chains [63]. Overall, probability distributions can be used for motion planning [64]- [66], but they usually do not strictly bound all possible future behaviors as required for provably safe motions.…”

Section: A Related Workmentioning

confidence: 99%

Set-Based Prediction of Traffic Participants Considering Occlusions and Traffic Rules

Koschi

Althoff

2021

IEEE Trans. Intell. Veh.

View full text Add to dashboard Cite

Provably safe motion planning for automated road vehicles must ensure that planned motions do not result in a collision with other traffic participants. This is a major challenge in autonomous driving, since the future behavior of other traffic participants is not known and since traffic participants are often hidden due to occlusions. In this work, we propose a formal setbased prediction that contains all acceptable future behaviors of both detected and potentially hidden traffic participants. Based on formalized traffic rules and nondeterministic motion models, we perform reachability analysis to predict the set of possible occupancies and velocities of vehicles, pedestrians, and cyclists. Real-world experiments with a test vehicle in various traffic situations demonstrate the applicability and real-time capability of our over-approximative prediction for both online verification and fail-safe trajectory planning. Even in congested, complex traffic scenarios, our forecasting approach enables self-driving vehicles to never cause accidents.

show abstract

“…Here, the hifi simulator of the previous case studies is now the lofi simulator. The new hifi simulator has a perception system 2 that uses LIDAR measurements to create a dynamic occupancy grid map (DOGMa) [19]- [21]. At each timestep, AST outputs the pedestrian acceleration and a single noise parameter, which is added to the distance reading of each beam that detects an object.…”

Section: Case Study: Perceptionmentioning

confidence: 99%

Finding Failures in High-Fidelity Simulation using Adaptive Stress Testing and the Backward Algorithm

Koren¹,

Nassar²,

Kochenderfer³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Validating the safety of autonomous systems generally requires the use of high-fidelity simulators that adequately capture the variability of real-world scenarios. However, it is generally not feasible to exhaustively search the space of simulation scenarios for failures. Adaptive stress testing (AST) is a method that uses reinforcement learning to find the most likely failure of a system. AST with a deep reinforcement learning solver has been shown to be effective in finding failures across a range of different systems. This approach generally involves running many simulations, which can be very expensive when using a high-fidelity simulator. To improve efficiency, we present a method that first finds failures in a low-fidelity simulator. It then uses the backward algorithm, which trains a deep neural network policy using a single expert demonstration, to adapt the low-fidelity failures to high-fidelity. We have created a series of autonomous vehicle validation case studies that represent some of the ways low-fidelity and highfidelity simulators can differ, such as time discretization. We demonstrate in a variety of case studies that this new AST approach is able to find failures with significantly fewer highfidelity simulation steps than are needed when just running AST directly in high-fidelity. As a proof of concept, we also demonstrate AST on NVIDIA's DriveSim simulator, an industry state-of-the-art high-fidelity simulator for finding failures in autonomous vehicles.

show abstract

Dynamic Environment Prediction in Urban Scenes using Recurrent Representation Learning

Cited by 32 publications

References 32 publications

Increasing the Efficiency of Policy Learning for Autonomous Vehicles by Multi-Task Representation Learning

Increasing the Efficiency of Policy Learning for Autonomous Vehicles by Multi-Task Representation Learning

Set-Based Prediction of Traffic Participants Considering Occlusions and Traffic Rules

Finding Failures in High-Fidelity Simulation using Adaptive Stress Testing and the Backward Algorithm

Contact Info

Product

Resources

About