Mobile Robot Navigation on Rugged and
Really Rugged Terrain
0800 - 1200, June 16
Professor Johann
Borenstein, University of Michigan - Ann Arbor
This workshop presents select topics of
mobile robotics research, presented in two distinct parts:
I. No GPS? No Problem!
This talk addresses the
problem of navigation on rugged, off-road terrain. After briefly addressing
issues of obstacle avoidance on 3-D terrain, the talk focuses most of its
attention on mobile robot position estimation in the absence of GPS. Unique
University of Michigan-developed methods for inertial navigation and for
detecting and correcting wheel slippage are presented. These methods were
originally developed for the Mars Rover 2009 mission, but they are applicable
to terrestrial dead-reckoning enhancements, as well. The talk concludes with
unique observations from the ultimate test for off-road mobile robots: The
DARPA Grand Challenge 2004.
II. Serpentine Robots - a
Can of Worms with Great Promise
This talk introduces the
relatively new breed of serpentine mobile robots. Serpentine robots, also
called Worm or Snake Robots, are elongated, multi-segmented vehicles that
promise to offer unprecedented mobility on extremely rugged terrain. The talk
reviews earlier work in this area, and then focuses on one state-of-the-art
serpentine robot, called "OmniTread." The OmniTread is described in detail and
provides the examples for a more in-depth discussion of serpentine
robots,their abilities, and their unique problems.
Back
Hybrid
Vehicle Power
Split Transmissions:
Exploring the electronic-CVT
0800 - 1200, June 16
Dr. John Miller, J-N-J
Miller, PLC
Abstract –
Gasoline and diesel electric hybrid propulsion
technologies, regardless of the source OEM, have selected the twin electric
machine power split configuration, known as electronic- CVT as the hybrid
transmission of choice. Early development of the e-CVT occurred when Toyota
undertook its project G21 in the mid-1990’s and then showcased the Prius
concept vehicle at the 1995 Tokoyo Auto Show that the electronic-CVT became
reality. In their quest for an economical and fuel efficient global vehicle
for the 21st century, Toyota Motor Corp. had its engineers focus on
power train architectures that would more than halve the CO2 emissions of
conventional gasoline power plant vehicles. Toyota is now in their second
generation of Toyota Hybrid System, THS, hybrid propulsion system technology.
The Ford Motor Company released its version of the e-CVT during the Fall of
2004 as the Escape SUV hybrid. The Ford Hybrid System, FHS, is an outgrowth
of earlier developments at the Volvo Car Company in conjunction with Aisin-AW,
a transmission subsidiary of the Toyota Motor Co. As an e-CVT, the FHS system
is a variant of the original THS-I power split with dual electric machines.
General Motors has independently developed a new type of power split referred
to as their Advanced Hybrid System-2 mode, or AHS-2, used in hybrid bus
propulsion by Allison for full size SUV applications. The AHS-2 system
realizes the input split character of the Toyota THS and Ford systems in its
Mode 1 configuration with extension to compound split in Mode 2
configuration. Transition between Mode 1, or low range for city driving, to
Mode 2, or high range for highway driving and towing, is accomplished during a
synchronous shift at the mechanical nodes of the system. In this seminar the
two main branches of power split architecture will be discussed in depth using
models and simulation.
Task-Centered Performance Analysis and Modeling
Professor Fatma Mili ,
Oakland University
1300 - 1700, June 16
The state of the art in
performance modeling is currently tightly bound to a specific task-agent
assignment. In other words, given a system under design such as a vehicle
(e.g. airplane), given a task under consideration (e.g. piloting), the
analysis of performance of this task starts once the task has been assigned to
agents (e.g. one human pilot). Performance models of the piloting task are
then developed around the constraints and assumptions of human cognitive,
perceptual, and motor architecture. This approach is suboptimal because the
assignment of tasks to agents is one of the key parameters in design.
Questions such as “how many people are needed to are best performance?” and
“Which agent would perform better on this task?” are important design
questions. Answering these questions is cumbersome and time consuming with
the current approaches to performance modeling. In this presentation, we will
describe a joint project whose goal is (i) to support reasoning about
performance in general, independently of the agent(s) performing the task, and
(ii) to support the assignment of agents to tasks. The approach is
task-centered: An OWL-S ontology is used to describe tasks and document their
needs in terms of resources. The resource consumption of a task can be given
as an estimate, or computed from the structure and the components of the task.
We describe the reasoning capability added to OWL-S to ensure that the
ontology is consistent ant to allow the resource consumption values to be
progressively refined as tasks are described in more detail.
Back
Tactically Aware Obstacle Avoidance and Path Planning
Alberto Lacaze,
Robotics Research
1300 - 1700, June 16
Several advance
technology areas for ground robotic need to be resolved for the success of the
Autonomous Navigation System (ANS). This tutorial will address four of these
topics.
1) Moving obstacle detection
from moving platform. Many algorithms that perform moving obstacle
detection from a static platform find their applicability in sentry
operations. However, the algorithms needed for moving obstacle detection and
tracking from a moving platform are different in principle. Moving obstacle
detection from a moving platform is an unavoidable requirement for ANS. Some
candidate systems will be presented.
2) Implementing tactically aware obstacles avoidance and path
planning. Tactics have been treated as a high level skill that does not
involve
obstacles avoidance and path planning. This assumption creates systems that
may be tactically aware in the long range plans; however, short
range behaviors are strongly influenced by the terrain (avoiding ditches and
vegetation). A system for integrating tactical behaviors and
obstacle avoidance will be presented together with the latest results of the
adaptable tactical behaviors competition.
3) Feature registration of the Leader-Follower operations.
Realistic scenarios for leader-follower operations will include GPS
shadows and INS (Inertial Navigation System) drift. On the other hand,
high speed leader-follower requirements under the ANS create restrictive
following error requirements. This presentation will concentrate on a visual
odometry techniques designed to meet these requirements. Aerial
ground-to-ground registration as well as air-to-ground registration will
be addressed.
4) Hardware acceleration for high speed obstacle avoidance. In the
past, stereo-based perception for ground autonomous vehicle was
the long pole in the tent of computational complexity. The shift in
popularity from stereo vision to LADAR-based systems as well as the
increased speed requirements for autonomous robots are focusing the
attention to the complexity of behavior generation. A methodology for
bringing hardware acceleration to the obstacle avoidance problem will be
presented.
Back
