Back to the Future: Robo Mobile 1995

On July 23, 1995, researcher Dean Pomerliau and Ph.D. Todd Jocke from the Institute of Robotics at Carnegie Mellon University set off on a journey through the states in an autonomous minivan. The computer program RALPH (Rapidly Adapting Lateral Position Handler, Fast Handler for Lateral Positioning, or BEAT) was used to determine the position on the road and the necessary actions to control the car. In the article, Jokem shares his memories of this journey and his assessments of the current state of roaming vehicles.

In recent years, robomobils have taken a strong place in the press. Projects published by Stanford University, Google, various automakers, Uber and Delphi. I perceived all the news about the “first robot” with pride and irony. Why? Because on July 23, 2015 it will be 20 years since the first test “Without Hands Through America” - the test of the first autonomous car. And I was lucky to be in a motley company from the Carnegie Melone Institute of Robotics, which built the car, and I was a passenger on a trip from Pittsburgh to San Diego.

Facts and figures

98.2% (2,797 of 2,849 miles) was traveled offline
')
Car: 1990 Pontiac Trans Sport Minivan

$ 20,000: software and hardware costs

Purpose of the trip: The Tonight Show With Jay Leno

It's amazing how much in the world of technology and social acceptance of autonomous cars has changed in 20 years, and at the same time it has not changed. I decided that it would be interesting to share thoughts from that time about our trip, and to understand what has changed and what remains the same.

Modern robomobili very stylish

I envy modern stylish cars from the world's leading automakers. We had a minivan with plastic door lining and fabric interior. But it was better than nothing. I would like to thank the visionary Ashok Ramaswami - this is a Delco company engineer (predecessor of Delphi), who discovered this Pontiac and donated it to us for experiments.

We build a system costing no more than $ 20k

We prepared the software and the car for four months and for an amount not exceeding $ 20,000. We had one computer based on 486DX2, a digital camera with a resolution of 640x480, a GPS receiver and a gyroscope.



It is interesting that GPS was used only to determine the speed. In those days, there was such a thing as GPS Selective Availability, that is, it was impossible to quickly and easily find out your exact location. And if you had found out, there were no suitable cards. But speed through GPS is better than nothing at all, and because of this we did not need to connect to the brains of the car.

How RALPH works

When the car is moving, the camera under the rear-view mirror removes the road, including markings, oil spots, curbs and traces of cars. The image is sent to the computer, which processes it and gives commands to the electric motor to turn the wheels to the left or to the right.

The system works on the PANS (Portable Advanced Navigation Support) platform, Extended Portable Navigation Support. It provides the calculation of information and data input / output, as well as the calculation of the position of the vehicle and control of the steering wheel. Powered by cigarette lighter.

More information about the operation of the RALPH system

Software problems

In 1994, Dean Pomerliau brought his software ALVINN, which was a neural network to track the position of the car in the strip, to such perfection that was possible. Anyway, of course, he had restrictions that did not allow him to use it at high speed on any kind of roads and in any weather.

One day, when he and his friend Chuck Thorpe drove a car in the area of ​​the Rocky Mountains in a blizzard, an insight came to him. It was so strong that he abandoned ALVINN and started making the system from scratch. And from January to May 1995, he built the RALPH system, which was not inferior in speed to ALVINN.



For full testing it was necessary to drive as much as possible. And then the plan was born to cross America under the ironic name “No Hands Across America” (based on the social action “Hands Across America”). We planned to take I-70 from Washington State to I-15 in Utah, and then turn onto San Diego.

From May to July, we honed the technology, planned stops, obtained permits and looked for sponsors. We did not find the madmen for the role of sponsor, so we did everything with small university money and with equipment borrowed from them.

The car was provided to Delco, and the GPS and gyro were given free of charge in exchange for a sponsor sticker on the car. To cover the cost of gasoline, accommodation and food, we sold T-shirts - for $ 10 apiece. Almost a system of sponsorship from NASCAR.

Great trip

Those interested can read the detailed travel journal . Suffice it to say that in these seven days we have learned more than the entire research community could have known in seven years. And have fun on the glory. We renewed the marriage vows in Vegas, saw a six-legged cow in Dogtown, crossed the Hoover Dam on the roaming mobile, met with Jem Leno - it was a great trip. One of the best moments was when the police stopped the journalist Otis Port, who was reporting on us, and at that time we drove past without holding the wheel.

Carnegie Melon University was the center of technical development of the romo mobiles and in the 80s and 90s people who moved this discipline came out of it. I was lucky to be in the right place at the right time and with the right people. Infrequently, life puts us in situations where we can do something that no one has done before. During the journey, we made discoveries and expanded the technical frontier. Now in this area, apart from these things, there are patent conflicts, concerns about security and problems with laws (but if you doubt about the technical discovery in this area - most likely, it was already done at the University from 1985 to 1997).