Tackling Occlusions & Limited Sensor Range with Set-based Safety Verification

Abstract

Provable safety is one of the most critical challenges in automated driving. The behavior of numerous traffic participants in a scene cannot be predicted reliably due to complex interdependencies and the indiscriminate behavior of humans. Additionally, we face high uncertainties and only incomplete environment knowledge. Recent approaches minimize risk with probabilistic and machine learning methods -even under occlusions. These generate comfortable behavior with good traffic flow, but cannot guarantee safety of their maneuvers. Therefore, we contribute a safety verification method for trajectories under occlusions. The field-of-view of the ego vehicle and a map are used to identify critical sensing field edges, each representing a potentially hidden obstacle. The state of occluded obstacles is unknown, but can be over-approximated by intervals over all possible states. Then set-based methods are extended to provide occupancy predictions for obstacles with state intervals. The proposed method can verify the safety of given trajectories (e.g. if they ensure collision-free fail-safe maneuver options) w.r.t. arbitrary safe-state formulations. The potential for provably safe trajectory planning is shown in three evaluative scenarios.

Occlusion-Aware Risk Assessment for Autonomous Driving in Urban Environments

“…Most closely related to the approach presented here are two risk quantification approaches [6], [7]. Orzechowski, Meyer, and Lauer [6] over-approximate all possible states of the incoming traffic by considering the leading edges of the visible polygon. Although safety is guaranteed, the resulting over-approximated polygons are not probabilistic, whereas our approach captures the full distribution of risk.…”

“…While the results in [7] look promising, they do not show how their risk assessment could be used for planning to achieve safer driving. Furthermore, both [6] and [7] show very limited results with only a single additional vehicle in the scene and it is unclear how these approaches perform in crowded scenes such as urban intersections. We focus on realistic intersections derived from real map data and occupied with many vehicles.…”

“…To the best of our knowledge, [6] is the only method that describes both risk assessment and planning in the same article. However, only a few examples were shown in their paper.…”

“…bilities by leveraging semantic and geometrical information of the environment, and anticipate the need to slow down due to the potential risk of collision that arises due to occlusions [6], [7]. In addition, earlier braking would reduce the maximum deceleration which consequently leads to greater ride comfort.…”

Behavior Planning of Autonomous Cars with Social Perception

“…where C R x t ,Î t , u R ; θ R is a cumulative cost over a horizon of N, as defined in (4). Note that with a long horizon N, discrete representation ofx t and b(x t ) is practically not feasible.…”