Ernst Dickmanns

from Wikipedia, the free encyclopedia

Ernst Dieter Dickmanns (born January 4, 1936 in Niederkassel ) is a German robotics scientist and university professor . Dickmanns was a professor at the University of the Federal Armed Forces in Munich (1975-2001) and is a pioneer of dynamic machine vision and autonomous vehicles . He was also visiting professor at CalTech in Pasadena and at the Massachusetts Institute of Technology , where he lectured on dynamic vision .

biography

Dickmanns was born in 1936. He studied aerospace engineering at the RWTH Aachen (1956-1961) and control engineering at Princeton University (1964-65). From 1961 to 1975 he worked at the German Aerospace Research and Research Institute (DFVLR) in Oberpfaffenhofen (today the German Aerospace Center , DLR) in the field of flight dynamics and flight path optimization. From 1971 to 1972 he worked as part of a Post-Doc Research Associateship at NASA - Marshall Space Flight Center in Huntsville on the reentry of space shuttles.

Pioneering work in autonomous driving

In the early 1980s , his team equipped a Mercedes-Benz van with cameras and other sensors. The 5-ton vehicle was modified so that the steering wheel , throttle valve and brakes could be controlled by a computer on the basis of real-time evaluation of image sequences. For safety reasons, the first experiments took place in Bavaria in 1986 on a disused airfield in Neubiberg and a motorway that had not yet been opened to public traffic. From 1987 the robot vehicle VaMoRs (test vehicle for autonomous mobility and computer vision) could drive completely autonomously at speeds of up to 96 km / h.

4-D approach

One of the greatest challenges in autonomous driving at high speed lies in the rapidly changing views of street scenes. At the time of Dickmanns' experiments, computers were much slower than they are today; therefore, sophisticated strategies for computer vision were necessary in order to be able to react in real time. Dickmanns' team solved the problem with an innovative approach to dynamic machine vision . From the beginning, spatiotemporal ( i.e. space-time) models were used, a so-called 4-D approach . Previous images did not necessarily have to be saved, but it was still possible to obtain estimates for all three-dimensional position and speed components. Attention control with artificial saccadic movements of the camera platform allowed the system to focus its attention on the most important details of a scene. Kalman filters were extended to the processing of perspective image sequences and were used to enable stable autonomous driving even in the presence of noise and uncertainty .

EUREKA program

In the years 1986/1987, the European research funding organization launched EUREKA at the suggestion of the European car industry, the project Program for a European Traffic of Highest Efficiency and Unprecedented Safety ( Prometheus ) were invested in the hundreds of millions of euros. The original plan of lateral routing through buried cables was quickly abandoned and replaced by the much more flexible machine vision approach, encouraged by Dickmanns' successes. Most of the larger automobile manufacturers took part in the program, Dickmanns and his team in collaboration with Daimler-Benz AG . Substantial progress was made in the following seven years: Dickmanns' robotic vehicles learned to move in traffic under different conditions. An accompanying human driver with a "red button" guaranteed that the vehicle could not get out of control and become a danger to the public. From 1992 onwards, driving on public roads was the norm. Several dozen transputers handled the enormous amount of calculations for the time.

PROMETHEUS project

Two high points were reached in 1994/1995 when Dickmanns' converted Mercedes-Benz S-Class vehicles completed international demonstrations. The first of these was the final presentation of the PROMETHEUS project in October 1994 on Autoroute 1 near Paris-Charles de Gaulle airport . The two vehicles VITA-2 from Daimler-Benz and VaMP ("VaMoRs Passenger Car" = test vehicle for autonomous mobility and computer vision) from UniBW Munich drove more than a thousand kilometers on the three-lane autobahn at speeds of up to 130 km / h . Driving in clear lanes, driving in a convoy with a speed-dependent safety distance and changing lanes to the left and right with automatic overtaking were demonstrated. For the overtaking maneuvers it was necessary to interpret the rear hemisphere as well. For this demonstration, two cameras with different focal lengths were used per hemisphere.

The second highlight was a 1758 km trip from Munich to Odense in Denmark to a project meeting and back in autumn 1995 . Both the longitudinal and the lateral control were carried out autonomously by computer vision. The robot vehicle reached speeds of over 175 km / h on the motorway . Publications from Dickmanns' research group indicate an autonomous driving distance without intervention of an average of 9 km; the longest autonomously driven section was 158 km. More than half of the interventions ( resets ) were carried out autonomously without human intervention. This is particularly noteworthy because the system only used black and white cameras and therefore could not model situations where yellow road markings on construction sites have priority over the white lines. During these journeys, 95% of the route was driven purely autonomously.

From 1994 to 2004, the older VaMoRs vehicle was used to develop driving skills on networks of smaller roads (including unpaved roads) and off-road, whereby obstacles such as potholes had to be avoided. Turning at intersections of unknown widths and different intersection angles presented itself as a great challenge, but it was finally achieved with the help of “expectation-based, multifocal, saccadic” vision (“expectation-based, multi-focal, saccadic vision”, EMS vision ) mastered. This vertebrate- derived vision mechanism uses animation skills based on known categories of moving objects (including the autonomous vehicle itself) and their possible behavior in certain situations. This wealth of experience was used to control the direction of gaze and attention.

aviation

The 4-D approach was tested not only in earthbound vehicles but also for dynamic vision of unmanned aerial vehicles (airplanes and helicopters). Autonomous sight land approaches and sight landings were demonstrated in hardware-in-the-loop simulations with a sensor fusion of sight and inertia data. Real landing approaches until shortly before touchdown could be demonstrated in 1992 with a two-propeller aircraft Do-128 from the University of Braunschweig at the airport there.

Another success of Dickmanns' approach to machine vision was the first successful vision-guided experiment to capture a free-floating object in weightlessness, which was carried out in 1993 on board the Columbia space shuttle as part of the D-2 Spacelab mission as part of the DLR 's' Rotex 'experiment was carried out.

Honors and prizes

See also

Individual evidence

  1. ↑ Dissolution of the acronym on Prof. Wünsches homepage ( Memento from March 4, 2016 in the Internet Archive )
  2. ^ Ernst D. Dickmanns: Dynamic Vision for Perception and Control of Motion . London 2007, ISBN 978-1-84628-637-7 .
  3. ↑ Fabian Hoberg, The man who taught cars to see , October 12, 2016 ( reference ).
  4. ↑ https://www.gfft-portal.de/verein/lösungen/ehrenverbindungen/

Web links

personal data
SURNAME Dickmanns, Ernst
ALTERNATIVE NAMES Dickmanns, Ernst Dieter
BRIEF DESCRIPTION German engineer
DATE OF BIRTH January 4, 1936
PLACE OF BIRTH Niederkassel