A Driving Simulator Interface for Micromobility
Abstract
Topic: Micromobility
For a number of years, the Elcano Project [1] has offered open-source electronics and software to drive small autonomous vehicles. The project has focused on electric tricycles, since these can move a person or goods in the city at the same speed as an automobile, but consume 45 times less energy [2]. Modest energy usage mean that a 10 kg battery is sufficient [3], which enables refueling by battery swap, eliminating range anxiety. A bank of batteries can be charged anytime solar or wind energy is available, providing an urban transportation system powered by renewable energy.
We have configured three prototype e-trikes in our lab, including one that is protected from weather by a light-weight body, weighs 75 kg, but is legally classified as a bicycle [4]. However, the project has not spread beyond the University of Washington, because of vehicle cost and the difficulty of mechanical conversion to an automated vehicle.
In this paper we present an electronic device that couples the microprocessors controlling vehicle sensors and actuators with an open-source driving simulator. This enables the e-trike control system to be tested on a virtual vehicle, which accelerates testing and development. The driving simulator is CARLA [5], which provides realistic vehicle physics under a wide range of street, weather and lighting conditions.
Other self-driving bicycle and tricycle projects have been done at MIT [6], Cornell [7], and India Institute of Technology [8]. Several companies have delivered goods by micro-vehicle [9]. The Intelligent Ground Vehicle Competition showcases small electric cars, but these typically weigh more than 500 kg [10]; we restrict micro-mobility to vehicles weighing less than 100 kg. A number of driving simulators are in active use, but most are proprietary. Most low-cost simulators are based on game physics rather than real-world physics.
Our system needs to provide a link between the CARLA simulator, which runs on a personal computer, and the microcontolers that drive the Elcano micro-vehicles. The Elcano design is a distributed system, running on a stack of Arduino microcontrollers linked over the automotive CAN bus [11]. The most important ones are the Drive-by-Wire system (Arduino Mega) and the navigation system (Arduino Due). The navigation system gathers information from various sensors, and combines them with a digital map to the destination. It outputs a desired heading and speed. The job of the Drive-by-Wire system is to make the actual course match the desired course. The Elcano System plugs each Arduino into a circuit board that connects it to the needed sensors or actuators.
The vehicle to simulator link is provided by a circuit board containing three Arduinos: Drive-by-Wire , Navigation, and Router. Instead of getting real sensor and actuator data, those sockets are connected to the Router which exchanges data with the CARLA PC.
The system has enabled us to control a virtual vehicle using the software written for the real vehicle. It has highlighted deficiencies in the existing code, and provided an expedited way for improvement.
A Driving Simulator Interface for Micromobility
Category
New Mobility Services
Description
Presenter: Tyler Folsom
Agency Affiliation: University of Washington, Bothell
Session: Technical Session C3: Micromobility for All?
Date: 6/1/2022, 3:30 PM - 5:00 PM
Presenter Biographical Statement: Tyler Folsom is a creative thinker and educator. Over the last 14 years, their main research interest has been automation of micro-mobility. They lead the open source Elcano Project supporting the transformation of electric trikes into Automated Vehicles (AV). Their research at the University of Washington, Bothell has led to three prototypes, with support from Amazon. They are the founder of Micro-AV SPC.
Dr. Folsom has applied skills in mathematics, software engineering and electrical engineering to tasks that include spacecraft control, metal cutting, in-process inspection of carbon composite layup, telemetry for flight test, reconditioning of nuclear reactors, and measurement of internal threads. Their involvement with robotic vehicles dates from the 2005 DARPA Grand Challenge of crossing the Mojave Desert autonomously. Dr. Folsom has been principal investigator for SBIR projects from the National Science Foundation, U.S. Air Force, Army and Navy.
They have taught a wide range of subjects in computer science and robotics at the Bothell campus of the University of Washington and DigiPen Institute of Technology. Dr. Folsom’s dissertation was Neural Networks Modelling Cortical Cells for Machine Vision where they invented an algorithm that can extract key information from the brain’s visual encoding.
Dr. Folsom’s curiosity has led to independent bicycle touring journeys on all inhabited continents, as well as backpacking trips into the wilderness. They devoted two and a half years to a bicycle trip around the world, visiting 49 countries.
Dr. Folsom is licensed as a Professional Engineer in the state of Washington. They have published 50 papers and technical reports, presenting the results at conferences in the United States, Europe, Asia and Canada. They wrote the e-book The Goddess at the Helm: Where Technology is Taking us, and how Activists can Change the Course.