Mobility goes towards electric or regenerative driving to achieve the goals of climate protection. Energy can be saved additionally by optimizing a velocity trajectory of the vehicles. For this purpose, a new driving assistance function has been developed to minimize the energy consumption of vehicles while mitigating the travel time and respecting some comfort criteria. Information about the route to be driven, map data (especially height profiles), GPS position as well as information about the surrounding traffic via sensors (camera, lidar, radar) are combined with detailed models of the vehicle and considered for the energy optimization. Typically, such kind of high dimensional optimization problems are solved using either heuristics or dynamic programming (DP) algorithms. In an embedded context, heuristics do not permit to exploit the full energy saving potential and are difficult to extend, while DP requires very high computing load and memory consumption and is therefore not implementable on µC-based control units. At Bosch we developed a very efficient holistic method to solve non-linear optimal control problems on µC-based control units. When applying this method to the eco-driving problem, runtime and memory consumption were significantly reduced, which enabled for the first time its implementation in the vehicle and a successful validation on the road.
Session: AUTONOMOUS DRIVING II | | 11:30 - 12:00