Safe driving requires autonomous vehicles to anticipate potential hidden traffic participants and other unseen objects, such as a cyclist hidden behind a large vehicle, or an object on the road hidden behind a building. Existing methods are usually unable to consider all possible shapes and orientations of such obstacles. They also typically do not reason about observations of hidden obstacles over time, leading to conservative anticipations. We overcome these limitations by (1) modeling possible hidden obstacles as a set of states of a point mass model and (2) sequential reasoning based on reachability analysis and previous observations. Based on (1), our method is safer, since we anticipate obstacles of arbitrary unknown shapes and orientations. In addition, (2) increases the available drivable space when planning trajectories for autonomous vehicles. In our experiments, we demonstrate that our method, at no expense of safety, gives rise to significant reductions in time to traverse various intersection scenarios from the CommonRoad Benchmark Suite.

Vehicle-to-everything (V2X) communication holds significant promise for augmenting autonomous driving capabilities. Particularly in dense traffic with occluded areas, V2X can be used to share information about the respective observed areas between traffic participants. In turn, reducing uncertainty about unseen areas can lead to less conservative behaviors while maintaining collision avoidance.This paper aims to leverage V2X to improve situation awareness for trajectory planning. We particularly address two challenges: First, the ego vehicle may not always receive up-to-date information. Second, some areas may remain occluded despite receiving information from other participants.In this work, we fuse the received information about the detected free space. We use reachability analysis to compute areas that are guaranteed to be free despite being occluded. This way, we can maintain collision-avoidance guarantees. We demonstrate the benefits of our proposed method both in simulations and physical experiments.

Autonomous agents rely on sensor data to construct representations of their environments, essential for predicting future events and planning their actions. However, sensor measurements suffer from limited range, occlusions, and sensor noise. These challenges become more evident in highly dynamic environments. This work proposes a probabilistic framework to jointly infer which parts of an environment are statically and which parts are dynamically occupied. We formulate the problem as a Bayesian network and introduce minimal assumptions that significantly reduce the complexity of the problem. Based on those, we derive Transitional Grid Maps (TGMs), an efficient analytical solution. Using real data, we demonstrate how this approach produces better maps than the state-of-the-art by keeping track of both static and dynamic elements and, as a side effect, can help improve existing SLAM algorithms.

Autonomous agents rely on accurate environment perception for safe and effective navigation. However, adverse weather conditions, such as snow or heavy rain, pose significant challenges to range sensors, leading to false positives and hindered obstacle detection. Previous work has focused on analyzing the sparsity of the measurements and learning from extensive datasets to filter out adverse weather detections. Despite promising results, the small residual unfiltered detections can still severely limit an agent's ability to generate feasible motion plans. In this paper, we introduce Transitional Grid Maps for Adverse Weather (TGMw). A temporal filter designed to distinguish weather-related noise from dynamic and static obstacles. It can either be used to directly filter the raw sensor measurements or to complement existing spatial filters. We demonstrate the benefits of our method in an extensive analysis in combination with multiple existing filters.