SwRI® Announces the Formation of the Active Vehicle Safety and Autonomy Consortium 

For more than 40 years, Southwest Research Institute (SwRI) has demonstrated broad capabilities developing integrated automated software, electronic, and communication systems to provide innovative and cost-effective solutions to transportation system problems.

In 2006, with its diverse expertise and state-of-the-art facilities and equipment, SwRI established the Southwest Safe Transport Initiative (SSTI), a $5 million internal research and development program, to improve safety in urban traffic environments. Through SSTI, the Institute is fusing the latest technology from multiple industries to meet the challenges associated with autonomous control of cars, trucks and tractors. SSTI is charged with developing new sensor, computing, and mobile technologies to augment vehicle platforms and provide autonomous vehicle capabilities.

SSTI draws on technologies and design methodologies from multiple industries such as:

• Unmanned aircraft systems
• Intelligent transportation systems
• Cognitive multi-agent systems
• Machine vision
• Engineering and vehicle dynamics
• Hardware/software-in-the-loop simulation
• Large-scale multi-function robotics
• Wireless communications
• Software engineering
• Safety and reliability systems
• Comprehensive test and evaluation of robotic systems

SwRI Autonomous Vehicle Development Platform

SwRI is developing a full-scale autonomous ground vehicle platform for advanced engineering of intelligent vehicle systems and applications development to support, both pre-competitive and competitive, research and development. Within the first six months of the program, SwRI engineers:

• Installed various prototype sensors targeted for future automotive applications and high performance blade computing hardware utilized for computationally intensive algorithms
• Developed extensive vehicle dynamic models to optimize control algorithms
• Established rudimentary software control of vehicle
• Optimized near real-time control algorithms allowing the SSTI vehicle to follow a dense set of waypoints at up to 50 km per hour around SwRI’s 1.87 km oval test track in San Antonio, Texas, without human intervention

Research is ongoing into sensor fusion, vehicle intelligence development and vehicle-to-vehicle communications.

AVSA Consortium

While autonomy is on the minds of vehicle manufacturers and suppliers, their primary concern is the development of active vehicle safety systems that warn the driver and can be implemented in production vehicle systems within the next five years. The benefits of an active collision avoidance system (which is the forerunner to fully autonomous driving) are undisputed, and OEMs and suppliers are challenged to stay abreast of this rapidly changing technology. It is becoming a difficult task for OEMs and suppliers to allocate sufficient in-house engineering resources to maintain a firm understanding of all the technology required for a vehicle to intelligently make decisions.

The Active Vehicle Safety and Autonomy Consortium was established by SwRI to advance this broad technology base, from a precompetitive standpoint, through collaborative research and development. By participating in the AVSA Consortium, OEMs and suppliers will minimize the increasing burden to engineering resources and be prepared to move this technology into their production systems on an accelerated schedule.

Objectives

SwRI will complete the initial phase of the SSTI program in early 2009 and intends to demonstrate advanced vehicle behaviors with the autonomous platform.

This platform will be available for the consortium research, as will the engineers and scientists who actively built the sensor fusion and intelligence systems. Pre-competitive research undertaken as part of AVSA-II might include:

• Vehicle behaviors and behavior prediction
• Intelligence and knowledge representation
• Frameworks for vehicle decision-making systems
• Cooperative vehicle maneuvers and interactions
• Development of advanced situational awareness methodologies

Proposed AVSA Consortium Projects

Structured Framework for Multiple Dynamic Object Classification:
Software that identifies and classifies objects that are provided by generic sensor systems, which could include vision, LIDAR, RADAR, GPS/INS, FLIR and others, will be developed. The framework will accept inputs from sensors in a variety of formats and external information sources such as satellite, Wi-Fi and DSRC. The framework will use updated sensor information about the environment to build an ontological structure through the inclusion of previously learned patterns of relationships. Both high-level and low-level sensor fusion techniques will be incorporated.

Sensor Fusion for Vehicle Detection and Classification:
Recognizing the difference between static obstacles such as highway barriers and dynamic obstacles such as other vehicles would be useful for on-road driving safety. Perhaps even more important than simply detecting a vehicle, however, is estimating its position and velocity. SwRI is developing algorithms to identify, classify and track multiple targets (in this case, street-legal vehicles) implementing high- and low-level sensor fusion techniques on information provided by both passive (such as cameras) and active (such as LIDAR, radar) sensors. Fusing data from multiple sources takes advantage of the strengths that both types of sensors provide. A detailed comparison of these strengths would be developed.

Creating a Learning Ontology for Situation Awareness:
Situation awareness is a key concept in the development of true autonomous vehicles and even active vehicle safety systems. Situational awareness is the ability to look at an environment and infer information based on the relationships between objects in that image and previously stored knowledge. In the past, vehicles have largely avoided the use of generic intelligence concepts by operating in highly constrained environments or with a safety driver. In the future, vehicles will have basic knowledge representation, situational awareness and the capability of learning when encountered objects are not represented in the knowledge base. The ultimate goal is to generate the ability to rapidly introduce new, usable information into the knowledge base and then allow this knowledge to propagate back down to the perception layer so that the newly classified objects can be identified directly (that is, in real time) the next time they are encountered.

Behavior Prediction and Behavior-based Decision Making:
Intelligent decision-making capabilities are becoming increasingly important for vehicle systems — for active safety systems now, and for vehicle autonomy in the future. The goal of this project is to investigate the use of behavior modeling and behavior identification in the decision-making process. Relevant sensor data, such as LIDAR, will be processed to identify vehicles and other static and dynamic objects. These will be combined with contextual information (relative location and speed) to identify possible behaviors, which will then be compared against the desired behaviors of the test vehicle to decide which one to enact.

Low Cost Navigation Sensor Software:
Navigation of both autonomous and human driving is dependent on good position/heading sensing in an urban setting. The cost of a GPS/INS system or a “laser range system” with high precision is currently too expensive for mass applications. Research is ongoing at SwRI to fuse basic sensor knowledge and situational awareness to enhance overall precision of location awareness in an urban setting. The goal of this project is to develop a low cost sensor, in software, to enhance precision and that operates reliably during GPS outages.

Collision Avoidance of Dynamic Objects:
Current collision avoidance systems focus on providing information to drivers, alerting them to entities that present a collision hazard. SwRI is conducting research that will develop the algorithms to detect and avoid collisions with objects that will influence the trajectory of the vehicle. This project will quantify threat potentials and establish early maneuvers that minimize the incidence of a collision. The proposed maneuver algorithms will use a rule-based architecture on the front end of the maneuver generation to decompose the problem into discrete events. In this case, the maneuvers are driven by the type of collision expected.

Determining Vehicle Control Thresholds in Active V2V and V2I Safety Systems:
Active safety systems that utilize communications can improve the safety of a vehicle. However, the point at which the vehicle takes over control from the driver is not well defined. This decision is based on numerous factors both inside and outside the vehicle. Developing the ability to dynamically detect vehicle control thresholds during — or preceding — activation of active safety systems and warnings allows the vehicle to respond appropriately and at the right time. This project would investigate a wide range of high risk situations to include left turn assist, lane changing, intersection collision avoidance and object proximity avoidance. Possible actions may be prevention of turning, reduction in vehicle speed or simply increasing driver alerts.

Relative Speed Detection Using Monocular Vision:
Nearly 40 percent of vehicle accidents and more than 20 percent of fatal vehicle accidents occur at intersections because the driver was not fully aware of the situation. Current programs at SwRI are developing vision processing algorithms to determine the velocity of objects at an intersection while the vehicle is stationary. The project proposes using optical flow and other image processing techniques to determine moving objects in consecutive frames of a video. This information enables technologies such as collision avoidance systems, pedestrian informers, object trackers and object classifiers.

Correlating Road Condition with Visual Cues:
Drivers instinctively know to avoid potholes and other rough road conditions because of past experiences. SwRI is providing autonomous vehicles the capability of identifying various detrimental road conditions using vision sensors. This project proposes developing an algorithm that uses an online training method where real road conditions (assessed using an accelerometer) are used to train a pattern matching system, such as a neural network, against images taken of the road patch before it is traversed.

Fuel Economy Enhancements Based on Situational Awareness:
Advances in informational technology, powertrain systems and geospatial mapping enables the vehicle control system to be aware of driving habits as well as route information ahead. Knowledge of upcoming grades, traffic situations and intersections can be used to control speed and gearing to get the best possible fuel economy. This project proposes the development of algorithms that enhance fuel economy by assisting the driver without compromising safety or perceived control. It is the goal of the research to quantify the fuel economy benefits from traffic conditions, intersections, grades and driver habit awareness.

Advanced Command and Control Algorithms:
Navigating a route autonomously, or in the event the driver is removed from an active safety maneuver, can be challenging even when a path can be charted. Tracking “heading” and “desired location” in a stable way requires robust control algorithm development. By combining optimum knowledge of vehicle state and control methodologies such as sliding mode control, SwRI is developing technology that contributes to Electronic Stability Control (ESC) mandated on all vehicles by 2012. This technology benefits autonomous vehicles as well. The goal is to improve safety and increase the precision of tracking a commanded route through the development of advanced control algorithms.

Robust System Condition Monitoring Tools for Reliable Active Safety and Autonomous Systems:
This project will develop a foundation for a robust system condition monitoring system. A roadblock to widespread adoption of advanced driver-assist and autonomous vehicle technologies is related to the perceived reliability for these systems. An independent system condition-monitoring algorithm must be able to detect and mitigate any anomalous behavior from sensor and state estimation vehicle control systems. This includes loss of communication, improper or unrecognized responses from sensors and others. A condition framework will be developed for a series of sensors or active control systems, which will describe the normal range of responses under various conditions and within response tolerances.

Benefits

AVSA leverages SwRI’s $5 million investment in the design and development of enabling technologies that enhance active safety and autonomous systems. R&D conducted as part of AVSA will begin with an advanced autonomous vehicle platform to allow the immediate execution of meaningful research maximizing the members’ return on investment.

Not a university driven activity, R&D is undertaken by professional engineers with multidisciplinary backgrounds. SwRI staff members are experts in numerous industries including unmanned aerial systems, intelligent transportation systems, cognitive multi-agent systems, machine vision, engineering dynamics, hardware/software-in-the-loop simulation, large-scale multifunction robotics, and safety and reliability systems. The pre-competitive research conducted by AVSA will leverage these capabilities, some of which the membership may not have in-house.

SwRI actively participates in numerous industry and standards bodies and will leverage the latest advancements in the execution of the AVSA research. The membership will benefit from this participation.

While other consortia, associations and joint projects operate in this arena, all require the participants to devote significant engineering resources in the execution of the R&D. AVSA removes this requirement and enables participants to benefit from pre-competitive R&D in future advanced technology and methodologies without having to reassign engineering staff away from the completion of near-term production systems.

Consortium Membership

AVSA I: Technical Information Exchange on SSTI Advanced Technology Development and Methodologies (September 2008-March 2009)

• Membership in AVSA I will afford members technical briefings at the member’s facilities where SwRI engineers will present technical methodologies and implementation specifics on the Institute’s Southwest Safe Transport Initiative.
• Membership in AVSA I will allow members to help shape the pre-competitive technology development that will be undertaken in AVSA II.

AVSA II: Pre-Competitive Technology Development (April 2009-March 2012)

• A yearly renewable contract is offered to members.
• The impact of the contribution is multiplied by the number of participants.
• SwRI’s internal research program involving autonomous vehicle enabling technology and methodologies will be shared with consortium members.
• SwRI will aggressively pursue patent applications for technology developed during the AVSA program.
• AVSA II members will receive a license to use the technology.
   

This brochure was published in April 2008. For more information about SwRI® Announces the Formation of the Active Vehicle Safety and Autonomy Consortium, contact Ryan Lamm, Manager, Phone (210) 522-5350, Fax (210) 522-2477, Automation and Data Systems Division, Southwest Research Institute, P.O. Drawer 28510, San Antonio, Texas 78228-0510.

Automation and Data Systems Division Brochures
SwRI Brochures SwRI Publications SwRI Home