AI-enabled fully autonomous systems

Self-driving cars, collaborative production assistants, and socially-enabled domestic robots are examples of autonomous systems at Bosch. Autonomous systems operate in complex and open-ended environments with high levels of independence and self-determination. They perceive, learn, reason and act with self-awareness and respond intelligently to unforeseen changes in the environment. We believe that such systems have the potential to deeply impact 21st century societies. They will increase quality of life for everyone, transform mobility, raise productivity, improve resource use, and enhance human safety.

In our research, we advance the state of the art for technologies critical to autonomous systems, particularly in computer vision, robotics, artificial intelligence, machine learning, optimization, control, validation and safety, sensors, and material science. In our basic research activities, we develop new theories and algorithms, collaborate with academia and follow an open innovation policy with our scientific publications and contributions to open source software. In applied research, we pre-develop and transfer research results to our business units. That allows Bosch to introduce new generations of intelligent products and services that bring innovation to hundreds of millions of people.

We focus on two areas of autonomous systems: automated driving and robotics.

Automated driving will be the cornerstone of future mobility systems. Highly automated and driverless cars will make traffic safer, more convenient and more efficient. Our research in automated driving spans a large variety of topics from deep learning for multi-modal scene perception and understanding, online decision-making, and large-scale simulations to novel methodologies for validation and safety. We additionally focus on training, testing, increasing robustness as well as generalization of our methods across different vehicles, use-cases, and business models. The outcomes affect partially and highly automated driving in fully automated and mixed traffic.

Robotics is both an enabling technology for autonomous systems and an application domain. We work on problems that extend the state of the art in simultaneous localization and mapping (SLAM), robot perception, planning, simulation and coordination and control, using both model and data-based methods. Examples include 24/7 year-round cm-accuracy life-long SLAM, long-term human motion prediction and deep reinforcement learning for manipulation. Our solutions are transferred into a uniquely diverse portfolio of products ranging from lightweight consumer robots, intralogistics transportation platforms, collaborative production assistants, and field robots for precision agriculture to 60-ton cargo movers.