The 2005 DARPA Grand Challenge – The Great Robot Race

Citation

(2008), "The 2005 DARPA Grand Challenge – The Great Robot Race", Industrial Robot, Vol. 35 No. 6. https://doi.org/10.1108/ir.2008.04935fae.001

The 2005 DARPA Grand Challenge – The Great Robot Race

Article Type: Book review From: Industrial Robot: An International Journal, Volume 35, Issue 6

Martin Buehler, Karl Iagnemma, Sanjiv Singh (Eds)Springer (www.springer.com) 2007£38.50520 pp.ISBN: 978-3-540-73428-4

The 2005 DARPA Grand Challenge was a 132 mile long race of autonomous ground vehicles through the Mojave Desert in October 2005. Of the 195 initial entrants, 43 teams successfully completed a preselection qualification round and participated in the National Qualification Event (NQE). After the NQE, the best 23 teams got the opportunity to participate in the race. The book presents 15 technical papers describing 16 of the 23 vehicles that participated in the DARPA Grand Challenge. It is not intended to be a textbook for graduate students, but is more an overview with a number of illustrative figures. It constitutes a good introduction to the area of autonomous vehicles, which covers computer vision, software integration, steering control, machinery, and obstacle avoidance. The chapters summarize the experiments and underlying system designs, although detailing the algorithms and presenting a failure analysis is outside the scope of most of the chapters in the book.

Chapter 1 “Stanley: the robot that won the DARPA ground challenge” specifies most parts of the winning system but does not discuss the system failures. On the other hand, Chapter 2, “A robust approach to high speed navigation for unrehearsed desert terrain”, details what happened when the system failed and it provides useful suggestions for future work. The next chapter describes “KAT-5: robust systems for autonomous vehicle navigation in challenging and unknown terrain”. The simplicity of this design improved the agility of the system. Furthermore, the heat of the desert was taken into consideration when developing the system. No details are given on what will happen if one of the LADAR devices is malfunctioning. Chapter 4, “The terramax autonomous vehicle” provides a simulation tool to evaluate the system’s performance prior to testing in a real environment, which can significantly reduce the cost. The next chapter, “Virginia Tech’s twin contenders: a comparative study of reactive and deliberative navigation”, introduces an approach to address the situation where GPS may fail and it provides a comparative study. The term “twin contenders” as used here is potentially confusing to some readers, in that it refers only to the presence of a backup vehicle and does not mean cooperation between two vehicles. Chapter 6, “Intelligent off-road navigation algorithms and strategies of team desert buckeyes in the DARPA grand challenge ’05”, details the algorithms that Ohio State applied and uses finite state machines to address complex situations. The next chapter, “The golem group/UCLA autonomous ground vehicle in the DARPA grand challenge”, presents interesting details of the algorithms used. No indication is given, however, how the 2005 version was improved as compared to the 2004 version.

Chapter 8, “CajunBot: architecture and algorithms”, reports about six different experiments and the electronic design. The authors introduce when the system failed and how to fix the problems. The next chapter, “SciAutonics-auburn engineering’s low cost high speed ATV for the 2005 DARPA grand challenge”, presents a summary of the team efforts for the race. Their system was able to detect and avoid obstacles with only half the number of sensors and computing power that was used by most of the competing teams. This efficient design could save cost in technology development. No information is given on the impact of sensor failure on system performance. In chapter 10, “Team CIMAR’s NaviGATOR: an unmanned ground vehicle for the 2005 DARPA grand challenge”, the authors used the JAUS model, which can improve the flexibility and compatibility of their system. In this case predefined data are needed to aid the system in navigation and readers may therefore be interested in the effect of poorly predefined data on performance. The next chapter, “Prospect Eleven: Princeton University’s entry in the 2005 DARPA grand challenge”, discusses how gaps in the terrain affect the system. This is important, and is perhaps not sufficiently discussed in the other chapters. Chapter 12 portrays “Cornell University’s 2005 DARPA grand challenge entry”. This chapter focuses exclusively on range sensors with a focus on how to evaluate and select the path for the robot. The next chapter, “A mixture-model based algorithm for real-time terrain estimation”, illustrates how to form a statistical representation of each sensor measurement in order to account for all sources of sensor error. This is significant in real-life cases and is detailed with several examples. Chapter 14 discusses “Alice: an information-rich autonomous vehicle for high-speed desert navigation”. The authors detail their system specification and in particular they discuss the no-data problem, which is important in real-life applications. The final chapter, “MITRE meteor: an off-road autonomous vehicle for DARPA’s grand challenge”, summarizes each of the system components and gives an overall design concept without a detailed technical explanation.

In summary, this book presents a nice collection of well-written chapters describing the overall system design of 16 of the vehicles that were competing in the DARPA Grand Challenge 2005. It is definitely worth reading.

Andreas KoschanUniversity of Tennessee, Knoxville, Tennessee, USA