Douglas Reece | Carnegie Mellon University (original) (raw)
Papers by Douglas Reece
Http Dx Doi Org 10 1162 105474603322391622, Mar 13, 2006
We have developed a movement behavior model for soldier agents who populate a virtual battlefield... more We have developed a movement behavior model for soldier agents who populate a virtual battlefield environment. Whereas many simulations have addressed human movement behavior before, none of them has comprehensively addressed realistic military movement at individual and unit levels. To design an appropriate movement behavior model, we found it necessary to elaborate all of the requirements on movement from the military tasks of interest, define a behavior architecture that encompasses all required movement tasks, select appropriate movement planning and control approaches in light of the requirements, and implement the planning and control algorithms with novel enhancements to achieve satisfactory results. The breadth of requirements in this problem domain makes simple behavior architectures inadequate and prevents any single planning approach from easily accomplishing all tasks. In our behavior architecture, a hierarchy of tasks is distributed over unit leaders and unit members. For movement planning, we use an A* search algorithm on a hybrid search space comprising a two-dimensional regular grid and a topological map; the plan produced is a series of waypoints annotated with posture and speed changes. Individuals control movement with reactive steering behaviors. The result is a system that can realistically plan and execute a variety of unit and individual agent movement tasks on a virtual battlefield.
Transportation Research Part A: Policy and Practice, 1993
Driving models are needed by many researchers to improve traffic safety and to advance autonomous... more Driving models are needed by many researchers to improve traffic safety and to advance autonomous vehicle design. To be most useful, a driving model must state specifically what information is needed and how it is processed. Such models are called computational because they tell exactly what computations the driving system must carry out. To date, detailed computational models have primarily been developed for research in robot vehicles. Other driving models are generally too vague or abstract to show the driving process in full detail. However, the existing computational models do not address the problem of selecting maneuvers in a dynamic traffic environment. In this paper we present a dynamic task analysis and use it to develop a computational model of driving in traffic. This model has been implemented in a driving program called Ulysses as part of our research program in robot vehicle development. Ulysses shows how traffic and safety rules constrain the vehicle's acceleration and lane use, and shows exactly where the driver needs to look at each moment as driving decisions are being made. Ulysses works in a simulated environment provided by our new traffic simulator called PHAROS, which is similar in spirit to previous simulators (such as NETSIM) but far more detailed. Our new driving model is a key component for developing autonomous vehicles and intelligent driver aids that operate in traffic, and provides a new tool for traffic research in general.
AI, Simulation, and Planning in High Autonomy …, 1994
Page 1. Extending DIS for Individual Combatants Douglas A. Reece Institute for Simulation and Tra... more Page 1. Extending DIS for Individual Combatants Douglas A. Reece Institute for Simulation and Training 1Jniversity of Central Florida 3280 Progress Drive Orlando, FL 32826 dreec:e@ rst.ucf.edu Abstract The domain of DIS ...
As mobile robots attempt more difficult tasks in more complex environments, they are faced with c... more As mobile robots attempt more difficult tasks in more complex environments, they are faced with combinatorially harder perceptual problems. In fact, computation costs for perception can easily dominate the costs for planning in a mobile robot. Existing perception systems on mobile robots are potentially many orders of magnitude too slow for real-world domains. In this paper we show active vision at the system level can make perception more tractable. We describe how our planning system for a complex domain, tactical driving, makes specific perceptual requests to find objects of interest. The perception system then scans the scene using routines to search for these objects in limited areas. This selective vision is based on an understanding and analysis of the driving task. We illustrate the effectiveness of request-driven routines by comparing the computational cost of general scene analysis with that of selective vision in simulated driving situations.
... DOCTORAL THESIS in the field of Computer Science Accesion Fort NTIS CRA&I - Selective Per... more ... DOCTORAL THESIS in the field of Computer Science Accesion Fort NTIS CRA&I - Selective Perception for Robot Driving DTIC TA8 DOUGLAS REECE J-stc--t . .... By Distributior Submitted in Partial Fulfillment of the Requirements Av. for the Degree of Doctor of Philosophy ...
Presence: Teleoperators and Virtual Environments, 2003
We have developed a movement behavior model for soldier agents who populate a virtual battlefield... more We have developed a movement behavior model for soldier agents who populate a virtual battlefield environment. Whereas many simulations have addressed human movement behavior before, none of them has comprehensively addressed realistic military movement at individual and unit levels. To design an appropriate movement behavior model, we found it necessary to elaborate all of the requirements on movement from the military tasks of interest, define a behavior architecture that encompasses all required movement tasks, select appropriate movement planning and control approaches in light of the requirements, and implement the planning and control algorithms with novel enhancements to achieve satisfactory results. The breadth of requirements in this problem domain makes simple behavior architectures inadequate and prevents any single planning approach from easily accomplishing all tasks. In our behavior architecture, a hierarchy of tasks is distributed over unit leaders and unit members. F...
![Research paper thumbnail of Synthetic soldiers [military training simulators]](https://attachments.academia-assets.com/70064486/thumbnails/1.jpg)
](IEEE Spectrum, 1997
ome two millennia ago, the historian Flavius Josephus explained the secret of the greatest army t... more ome two millennia ago, the historian Flavius Josephus explained the secret of the greatest army the world had ever seen "The Romans are certain of victory., .because their exercises are
ROAD USER BEHAVIOUR: THEORY AND …, 1988
... Title: AN OVERVIEW OF THE PHAROS TRAFFIC SIMULATOR. Accession Number: 00445191. Language: Eng... more ... Title: AN OVERVIEW OF THE PHAROS TRAFFIC SIMULATOR. Accession Number: 00445191. Language: English. ... Order URL: http://library.its.berkeley.edu. Index Terms: Automatic control; Automobiles; Pharos computer program; Robotics; Simulation; Traffic flow. Subject Areas: ...
Proceedings of the 2007 …, 2007
Page 1. Asymmetric Adversary Tactics for Synthetic Training Environments Brian S. Stensrud, Dougl... more Page 1. Asymmetric Adversary Tactics for Synthetic Training Environments Brian S. Stensrud, Douglas A. Reece, Nicholas Piegdon Soar Technology, Inc. 3361 Rouse Road, Suite #175, Orlando, FL 32817 {stensrud, douglas.reece, piegdon}@soartech.com ...
Computer Science Dept., Carnegie-Mellon University, Pittsburgh, PA, 1990
Computer vision research aimed at performing general scene understanding has proven to be concept... more Computer vision research aimed at performing general scene understanding has proven to be conceptually difficult and computationally complex. Active vision is a promising approach to solving this problem. Active vision systems use optimized sensor settings, reduced fields of view, and relatively simple algorithms to efficiently extract specific information from a scene. This approach is only appropriate in the context of a task that motivates the selection of the information to extract. While there has been a fair amount of research that describes the extraction processes, there has been little work that investigates how active vision could be used for a realistic task in a dynamic domain. We are studying such a task: driving an autonomous vehicle in traffic. In this paper we present a method for controlling visual attention as part of the reasoning process for driving, and analyze the efficiency gained in doing so. We first describe a model of driving and the driving environment, and estimate the complexity of performing the required sensing with a general driving-scene understanding system. We then introduce three programs that use increasingly sophisticated perceptual control techniques to select perceptual actions. The first program, called Ulysses-l, uses perceptual routines, which use known reference objects to guide the search for new objects. The second program, Ulysses-2, creates an inference tree to infer the effect of uncertain input data on action choices, and searches this tree to decide which data to sense. Finally, Ulysses-3 uses domain knowledge to reason about how dynamic objects will move or change over time; objects that do not move enough to affect the robot's decisions are not selected as perceptual targets. For each technique we have run experiments in simulation to measure the cost savings realized by using selective perception. We estimate that the techniques included in Ulysses-3 reduce the computational cost of perception by 9 to 12 orders of magnitude when compared to a general perception system.
Http Dx Doi Org 10 1162 105474603322391622, Mar 13, 2006
We have developed a movement behavior model for soldier agents who populate a virtual battlefield... more We have developed a movement behavior model for soldier agents who populate a virtual battlefield environment. Whereas many simulations have addressed human movement behavior before, none of them has comprehensively addressed realistic military movement at individual and unit levels. To design an appropriate movement behavior model, we found it necessary to elaborate all of the requirements on movement from the military tasks of interest, define a behavior architecture that encompasses all required movement tasks, select appropriate movement planning and control approaches in light of the requirements, and implement the planning and control algorithms with novel enhancements to achieve satisfactory results. The breadth of requirements in this problem domain makes simple behavior architectures inadequate and prevents any single planning approach from easily accomplishing all tasks. In our behavior architecture, a hierarchy of tasks is distributed over unit leaders and unit members. For movement planning, we use an A* search algorithm on a hybrid search space comprising a two-dimensional regular grid and a topological map; the plan produced is a series of waypoints annotated with posture and speed changes. Individuals control movement with reactive steering behaviors. The result is a system that can realistically plan and execute a variety of unit and individual agent movement tasks on a virtual battlefield.
Transportation Research Part A: Policy and Practice, 1993
Driving models are needed by many researchers to improve traffic safety and to advance autonomous... more Driving models are needed by many researchers to improve traffic safety and to advance autonomous vehicle design. To be most useful, a driving model must state specifically what information is needed and how it is processed. Such models are called computational because they tell exactly what computations the driving system must carry out. To date, detailed computational models have primarily been developed for research in robot vehicles. Other driving models are generally too vague or abstract to show the driving process in full detail. However, the existing computational models do not address the problem of selecting maneuvers in a dynamic traffic environment. In this paper we present a dynamic task analysis and use it to develop a computational model of driving in traffic. This model has been implemented in a driving program called Ulysses as part of our research program in robot vehicle development. Ulysses shows how traffic and safety rules constrain the vehicle's acceleration and lane use, and shows exactly where the driver needs to look at each moment as driving decisions are being made. Ulysses works in a simulated environment provided by our new traffic simulator called PHAROS, which is similar in spirit to previous simulators (such as NETSIM) but far more detailed. Our new driving model is a key component for developing autonomous vehicles and intelligent driver aids that operate in traffic, and provides a new tool for traffic research in general.
AI, Simulation, and Planning in High Autonomy …, 1994
Page 1. Extending DIS for Individual Combatants Douglas A. Reece Institute for Simulation and Tra... more Page 1. Extending DIS for Individual Combatants Douglas A. Reece Institute for Simulation and Training 1Jniversity of Central Florida 3280 Progress Drive Orlando, FL 32826 dreec:e@ rst.ucf.edu Abstract The domain of DIS ...
As mobile robots attempt more difficult tasks in more complex environments, they are faced with c... more As mobile robots attempt more difficult tasks in more complex environments, they are faced with combinatorially harder perceptual problems. In fact, computation costs for perception can easily dominate the costs for planning in a mobile robot. Existing perception systems on mobile robots are potentially many orders of magnitude too slow for real-world domains. In this paper we show active vision at the system level can make perception more tractable. We describe how our planning system for a complex domain, tactical driving, makes specific perceptual requests to find objects of interest. The perception system then scans the scene using routines to search for these objects in limited areas. This selective vision is based on an understanding and analysis of the driving task. We illustrate the effectiveness of request-driven routines by comparing the computational cost of general scene analysis with that of selective vision in simulated driving situations.
... DOCTORAL THESIS in the field of Computer Science Accesion Fort NTIS CRA&I - Selective Per... more ... DOCTORAL THESIS in the field of Computer Science Accesion Fort NTIS CRA&I - Selective Perception for Robot Driving DTIC TA8 DOUGLAS REECE J-stc--t . .... By Distributior Submitted in Partial Fulfillment of the Requirements Av. for the Degree of Doctor of Philosophy ...
Presence: Teleoperators and Virtual Environments, 2003
We have developed a movement behavior model for soldier agents who populate a virtual battlefield... more We have developed a movement behavior model for soldier agents who populate a virtual battlefield environment. Whereas many simulations have addressed human movement behavior before, none of them has comprehensively addressed realistic military movement at individual and unit levels. To design an appropriate movement behavior model, we found it necessary to elaborate all of the requirements on movement from the military tasks of interest, define a behavior architecture that encompasses all required movement tasks, select appropriate movement planning and control approaches in light of the requirements, and implement the planning and control algorithms with novel enhancements to achieve satisfactory results. The breadth of requirements in this problem domain makes simple behavior architectures inadequate and prevents any single planning approach from easily accomplishing all tasks. In our behavior architecture, a hierarchy of tasks is distributed over unit leaders and unit members. F...
IEEE Spectrum, 1997
ome two millennia ago, the historian Flavius Josephus explained the secret of the greatest army t... more ome two millennia ago, the historian Flavius Josephus explained the secret of the greatest army the world had ever seen "The Romans are certain of victory., .because their exercises are
ROAD USER BEHAVIOUR: THEORY AND …, 1988
... Title: AN OVERVIEW OF THE PHAROS TRAFFIC SIMULATOR. Accession Number: 00445191. Language: Eng... more ... Title: AN OVERVIEW OF THE PHAROS TRAFFIC SIMULATOR. Accession Number: 00445191. Language: English. ... Order URL: http://library.its.berkeley.edu. Index Terms: Automatic control; Automobiles; Pharos computer program; Robotics; Simulation; Traffic flow. Subject Areas: ...
Proceedings of the 2007 …, 2007
Page 1. Asymmetric Adversary Tactics for Synthetic Training Environments Brian S. Stensrud, Dougl... more Page 1. Asymmetric Adversary Tactics for Synthetic Training Environments Brian S. Stensrud, Douglas A. Reece, Nicholas Piegdon Soar Technology, Inc. 3361 Rouse Road, Suite #175, Orlando, FL 32817 {stensrud, douglas.reece, piegdon}@soartech.com ...
Computer Science Dept., Carnegie-Mellon University, Pittsburgh, PA, 1990
Computer vision research aimed at performing general scene understanding has proven to be concept... more Computer vision research aimed at performing general scene understanding has proven to be conceptually difficult and computationally complex. Active vision is a promising approach to solving this problem. Active vision systems use optimized sensor settings, reduced fields of view, and relatively simple algorithms to efficiently extract specific information from a scene. This approach is only appropriate in the context of a task that motivates the selection of the information to extract. While there has been a fair amount of research that describes the extraction processes, there has been little work that investigates how active vision could be used for a realistic task in a dynamic domain. We are studying such a task: driving an autonomous vehicle in traffic. In this paper we present a method for controlling visual attention as part of the reasoning process for driving, and analyze the efficiency gained in doing so. We first describe a model of driving and the driving environment, and estimate the complexity of performing the required sensing with a general driving-scene understanding system. We then introduce three programs that use increasingly sophisticated perceptual control techniques to select perceptual actions. The first program, called Ulysses-l, uses perceptual routines, which use known reference objects to guide the search for new objects. The second program, Ulysses-2, creates an inference tree to infer the effect of uncertain input data on action choices, and searches this tree to decide which data to sense. Finally, Ulysses-3 uses domain knowledge to reason about how dynamic objects will move or change over time; objects that do not move enough to affect the robot's decisions are not selected as perceptual targets. For each technique we have run experiments in simulation to measure the cost savings realized by using selective perception. We estimate that the techniques included in Ulysses-3 reduce the computational cost of perception by 9 to 12 orders of magnitude when compared to a general perception system.