<p style="text-align: justify">Perception is one of the primary tasks in autonomous driving. Exteroceptive measurement sensors like lidar, radar, and camera perceive the autonomous vehicle's surrounding environment. Also, vehicles can share events-related information through communications with other vehicles or infrastructure with connectivity. Therefore, a connected autonomous vehicle can inform other nearby vehicles of their surrounding conditions that are not within their field of view. The connected vehicles can receive information about anticipated or forthcoming conditions and uncertainties ahead. This communication-enabled sensing shared between surrounding vehicles is called cooperative perception. The ego vehicle perceives forthcoming conditions by augmenting its field of view virtually through receiving useful scene data from surrounding vehicles.&nbsp; These data alert the ego vehicle of possible critical conditions or hazards ahead. Thus, the ego vehicle can plan its near-term trajectory accordingly.&nbsp; Cooperative perception extends the autonomous vehicles' sensing capabilities. It can help a self-driving vehicle drive safely and prevent accidents caused by undetected vulnerable road users (VRUs) due to sensor blockage or insufficient sensory field of view, especially within two to three times the braking distances.</p><p style="text-align: justify">For this connected, cooperative perception between two vehicles to work correctly, the two vehicles' relative positions need to be established through robust and accurate estimation.&nbsp; This paper first focuses on relative pose estimation for two vehicles' cooperative perception-based autonomous driving through direct and indirect methods. The direct method involves the sensors' data covering the surrounding area under both vehicles' common field of view. The indirect method utilizes the positioning data (GPS) of each vehicle. The indirect relative pose estimation method is highly challenging where uncertainties are often involved due to full or partial occlusions caused by dynamic or moving objects or VRUs. Hence the direct method will help to anticipate relative pose fixing. After that, the study also focuses on tracking multiple moving objects, especially vulnerable road users, by combining the sensors of two autonomous vehicles connected in the network and sharing the objects' useful location information.&nbsp; The objective is to enhance each autonomous vehicle's perception capability beyond its own respective field of view.&nbsp; This proposed method estimates the VRUs' positions and velocities based on a Probability Hypothesis Density Filter, a multi-target Bayesian filter to overcome the data association problem. Finally, moving objects' position and velocity information will be shared between two vehicles for their future safe path planning to avoid hazardous situations. Therefore, the method will dynamically handle the specific or uncertain hazards that are otherwise unavoidable. The method is tested using simulations and an emulated mobile robot environment depicting real traffic scenarios.</p>

Relative Pose Estimation for Cooperative Vehicles and Tracking of Multiple Dynamic Objects

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.