Robust and embeddable optimal control techniques have revolutionized the automotive industry in the last decades in at least two domains: safety and energy. Both domains have challenged control engineers with with complex problems, where advanced and nonlinear control techniques needed to be implemented in inexpensive automotive-grade hardware.
We have helped customers solving such problems via providing our expertise in modelling, problem formulation and efficient hardware implementation. The example list below is not exhaustive but only representative of the main areas of recent interest.
The three pillars of autonomous driving are sensing, localization & mapping, and motion planning. Embotech provides bespoke solutions for the motion planning problem as well as offering specific products (PRODRIVER) for it. Using our core technology FORCESPRO as a basis, we work with partners an end-users to develop solutions which address the level of autonomous driving and certification required. Our solutions are highly effective on automotive-grade embedded systems. Solutions are model-based, certifiable, easy to integrate and quick to market.
The best performing vehicle motion control techniques are able to handle constraints and guarantee feasibility for long prediction horizons. Model predictive control (MPC), a robust and well understood method, is among the most common such techniques. MPC requires the online solution of an optimization problem that needs to be solved in real time in inexpensive embedded hardware. Our core technology, FORCESPRO, has been designed to solve such problem efficiently. A customer-tailored motion controller based on FORCESPRO is one of our most popular solution.
Automating logistic yards will be one of the earliest applications of autonomous driving. This is because safety issues are more easily mitigated by the highly controllable environment and by the centralization of the perception system.
We have worked with customers to design a solution that satisfies their particular requirements. Typical implementations include infrastructure-based sensing and cloud computing. This allows the centralization of the perception systems and the environment modelling, and the coordination of vehicles via a secure link. Vehicles don’t need to have sophisticated sensing, but only the ability to execute a motion control algorithm as well as having drive-by-wire capabilities.
Due to the technological and physical limitations of current energy storage technologies, driving range is one of the biggest challenges when it comes to public acceptance of electric propulsion systems.
Our energy management solutions, developed for purely electric but also for hybrid-electric powertrains, makes use of state-of-the-art numerical optimization technologies in order deliver the optimum range out of the given powertrain and energy storage setup. It seamlessly optimizes the velocity profile, the battery state-of-charge and, in the case of hybrid powertrains, the power distribution between combustion engine and electric motor.
- Builds on state-of-the-art real-time optimization technology
- Delivers smooth driving trajectories – passenger comfort
- Customizable based on the driving style of the customer
- Easily adapted for any vehicles – no machine learning or vehicle training for motion planning
- Very close to an offline dynamic programming solution
- Achieves close to global energy optimal driving in real-world scenarios
- Delivering on a ~5km horizon and an update rate of ~100ms
- Uses non-linear maps of the powertrain components
- Can include complex constraint
Sophisticated automotive active safety system (stability control, collision avoidance, etc.) rely on fast feedback control loops. Model predictive control, when implemented correctly, can dramatically improve and enhance the performance of active safety systems. Since a solution of an online (often nonlinear) optimization problem is also required for such applications, FORCESPRO can also in this case show superior performance.
We have been helping customers develop active safety solutions such as torque vectoring and active suspension systems based on real-time implementation of embedded MPC.
race assist and autonomous racing
A niche but growing market is the one of autonomous racing and race assist (feeding back to a human driver the real-time calculated optimal trajectories). Our FORCESPRO based motion planners are easily tuned to race mode, where minimization of time is given more importance than comfort in the underlying optimization problem.
As an application example, the Formula Student Autonomous team of ETH Zurich (AMZ racing) has successfully won competitions by using our motion planner. The video on the left shows how the motion planner makes prediction and generates trajectories continuously and in real time.
You may be interested in knowing that the winning car is physically located in our office in Zurich. It’s worth paying a visit to see it and to find out more.
Looking for more information?
If you’d like to find out more our Automotive Solutions, let us know. Either use the webform on the right or book a meeting via the link below. Our team of experts will be happy to discuss more details with you.