Latest News / Dec ‘23 / Amazon Web Services (AWS) joined the Autoware Foundation as a premium member!

Embotech

Embotech is a leading developer of cutting-edge decision-making software. Our embeddable software empowers autonomous systems to make decisions by solving complex optimization problems in milliseconds, bringing significant improvements in safety, productivity and energy efficiency. Current applications include Passenger and commercial vehicles and Industry and Robotics Solutions.

Value Proposition

Embotech provides solutions to autonomous navigation, path planning, tracking and control for autonomous or highly-automated vehicles and vehicle fleets. Our product, PRODRIVER, generates drivable trajectories or direct actuator commands such as steering, accelerating and braking. These are computed given the surrounding environment information. PRODRIVER does so by continuously making predictions and solving an optimization problem in real-time.

Technology Description

Our design philosophy focuses on a virtual ground vehicle driver that drives like a human but does not make human-like mistakes. We avoid learning-based methods or hard-coded rule-based heuristics where either repeatability and determinism cannot be guaranteed, and a large amount of training data is required, or where it’s impossible to encode all situations and the trajectories tend to feel “artificial”. We don’t believe in what we call “situational pre-programming”.

The problem of driving is complex, but mathematical functions can describe it. Such functions are the equations of motions of the vehicle, the tire dynamics and the constraints imposed by physics (e.g. maximum turning radius), regulation (e.g. speed limits) and safety (e.g. obstacles, road conditions). Those are the ingredients required to set up the mathematical optimization problem.
We take up the challenge to solve the problem in real-time – within milliseconds.

The advantage is that the trajectories generated by PRODRIVER are smooth, human-like, consistent, and able to deal with scenarios never encountered before. By modifying a few parameters, different driving styles automatically emerge. For example, if, in the cost function of the optimization problem, we try to minimize lateral accelerations, we observe a driving style similar to that of a careful chauffeur. On the other hand, if the aim is to minimize time, the result is a more race-like style.
Most importantly, our trajectories are physically consistent and correct: the vehicle can follow them. This also allows us to deal with driving on low-friction surfaces like mud, gravel, snow, and ice and to create emergency avoidance manoeuvres at the limits of handling within the same piece of software.