## Theory and Experiments in Autonomous Sensor-Based Motion Planning with Applications for Flight Planetary Microrovers (1999)

Citations: | 16 - 3 self |

### BibTeX

@TECHREPORT{Laubach99theoryand,

author = {Sharon Lynn Laubach},

title = {Theory and Experiments in Autonomous Sensor-Based Motion Planning with Applications for Flight Planetary Microrovers},

institution = {},

year = {1999}

}

### OpenURL

### Abstract

With the success of Mars Pathfinder's Sojourner rover, a new era of planetary exploration has opened, with demand for highly capable mobile robots. These robots must be able to traverse long distances over rough, unknown terrain autonomously, under severe resource constraints. Much prior work in mobile robot path planning has been based on assumptions that are not truly applicable to navigation through planetary terrains. Based on the author's firsthand experience with the Mars Pathfinder mission, this work reviews issues which are critical for successful autonomous navigation of planetary rovers. No current methodology addresses all of these constraints. We next develop the sensor-based "Wedgebug" motionplanning algorithm. This algorithm is complete, correct, requires minimal memory for storage of its world model, and uses only on-board sensors, which are guided by the algorithm to e#ciently sense only the data needed for motion planning, while avoiding unnecessary robot motion. The p...

### Citations

2686 |
A robust layered control system for a mobile robot
- Brooks
- 1986
(Show Context)
Citation Context ...e complete knowledge of the environment beforehand, and that often the environment must be bounded to guarantee completeness. The class of heuristic planners, such as Brooks’ subsumption architectur=-=e [5] or -=-the track arbitration schemes developed at CMU [26],[78], as well as the “Go To Waypoint” algorithm employed by the Sojourner rover [54] and the Rocky7 rover [82] (see Table 1.1), share the useful... |

1995 |
Robot Motion Planning
- Latombe
- 1991
(Show Context)
Citation Context ...y desirable virtue: the planner will generate a path if one is possible, and will halt otherwise in finite time. Latombe describes classical planners in some detail in his book, Robot Motion Planning =-=[36], in w-=-hich he splits the classical planners (a) roadmap (b) exact cell decomposition (c) approx. cell decomposition (d) potential field Figure 1.1: Examples of types of “classical” planners ∗The 2D wo... |

525 |
Morse Theory
- Milnor
- 1963
(Show Context)
Citation Context ...ry (described in detail in [17]). We will not discuss this theory in detail, but rather note that in this case, it can be shown that fT is a Morse function on ∂Oi, given a generic environment ‡ [5=-=5], [51]. One-=- nice property of Morse functions is that their critical points are non-degenerate § , and therefore isolated [18], [55]. Since Morse functions are smooth, it follows that V ∗ is continuous on ∂O... |

481 |
The Complexity of Robot Motion Planning
- Canny
- 1988
(Show Context)
Citation Context ...navigation (see Fig. 1.1(a)). Examples of roadmap algorithms include the visibility and reduced visibility (also known as tangent) graphs [57],[36], Voronoi graphs [58], and Canny’s silhouette metho=-=d [7]. Ea-=-ch of these schemes constructs a set of one-dimensional curves which encapsulate the topology of the free space, and serve as a system of “freeways,” complete with methods for entry and exit, to e... |

340 | Spatial Planning: A Configuration Space Approach
- Lozano-Pérez
- 1983
(Show Context)
Citation Context ...en extend the proof’s domain of applicability. 3.2 Concepts from Motion Planning In his book, Robot Motion Planning [36], Latombe succinctly describes the notion (originally developed by Lozano-Pér=-=ez [39]-=-) of a configuration space, C. The underlying concept is to represent the real-world robot as a point in an appropriate space, and to map obstacles into this same space. Then, the space contains a con... |

335 |
Robot motion planning: A distributed representation approach
- Barraquand, Latombe
- 1991
(Show Context)
Citation Context ...991) [12], utilises as10 heuristic planner based on Slack’s “navigation templates” [75] for obstacle avoidance and goal seeking, as well as a version of Barraquand and Latombe’s non-holonomic =-=planner [4] to -=-recover from cul-de-sacs. The planner was implemented on JPL’s Robby, a large prototype rover. This planner is claimed to be complete, though to the author’s knowledge, no proof has been presented... |

301 |
Exact robot navigation using artificial potential fields
- Rimon, Koditschek
- 1992
(Show Context)
Citation Context ...nal formulation of the potential field method did not possess the property of completeness [27], further development of the theory has produced a classical potential field algorithm which is complete =-=[66]-=-. As previously stated, the classical planners as a group possess the key advantages of provable completeness and correctness. Furthermore, since the world model is known a priori, allowing these algo... |

264 |
Differential Topology
- Guillemin, Pollack
- 1974
(Show Context)
Citation Context ...s will be shown later † Specifically, Kamon, Rimon, and Rivlin state that there are a finite number of local minima in the (assumed bounded) free configuration space.s59 (for differential geometry) =-=, [18] and-=- [19] (for topology). A brief “roadmap” of the remainder of this section follows: After establishing the basic properties of the relevant space, we prove the proposition that all of the local mini... |

211 | The focussed D ∗ algorithm for real-time replanning
- Stentz
- 1995
(Show Context)
Citation Context ...t often the environment must be bounded to guarantee completeness. The class of heuristic planners, such as Brooks’ subsumption architecture [5] or the track arbitration schemes developed at CMU [26=-=],[78], as-=- well as the “Go To Waypoint” algorithm employed by the Sojourner rover [54] and the Rocky7 rover [82] (see Table 1.1), share the useful property of being able to be made sensorbased much more eas... |

194 |
Stratified Morse Theory
- Goresky, MacPherson
- 1988
(Show Context)
Citation Context ...roes at T and where V is normal to the obstacle boundary. Since the obstacle boundaries are smooth, we have a situation applicable for the machinery of stratified Morse theory (described in detail in =-=[17]). W-=-e will not discuss this theory in detail, but rather note that in this case, it can be shown that fT is a Morse function on ∂Oi, given a generic environment ‡ [55], [51]. One nice property of Mors... |

179 |
S.: Differential Equations, Dynamical Systems, and Linear Algebra
- Hirsch, Smale
- 1974
(Show Context)
Citation Context ...; otherwise, Z�{c} ⊆Bx, where {c} is the set of (at most two) y - 2 x 2 y - 3 x 3 ] ts68 “corner points” in Z. With some abuse of language, we call the interval Bx the basin of attraction of x=-= for fT [20]. In-=-tuitively, if we were to let a particle follow the (negative) gradient of fT |Cj , starting at any point in Bx, the particle would end at x (or on an arc Ax containing x such that fT(y)=fT(x)∀y∈Ax... |

178 | Optimal and efficient path planning for partially-known environments
- Stentz
- 1994
(Show Context)
Citation Context ...thm, developed by Kamon, Rivlin, and Rimon [23]. The second approach springs from approximate cell decomposition, filling in a grid-based world model incrementally, such as Stentz’ D* algorithm [78]=-=, [77]. Th-=-e third approach harks more closely to the heuristic planners, and includes the “Bug” algorithms of Lumelsky and Stepanov [40], [64], which combine reactive behaviours with global parameters to re... |

151 |
Introductory Real Analysis
- Kolmogorov, Fomin
- 1970
(Show Context)
Citation Context ...he boundaries are continuous and rectifiable.) We assume that the reader has basic knowledge of real analysis, differential geometry, and of topology; good references for these topics are [69], [68], =-=[29] (fo-=-r real analysis), [31] ∗ or rather basins of attraction, as will be shown later † Specifically, Kamon, Rimon, and Rivlin state that there are a finite number of local minima in the (assumed bounde... |

149 | Path-planning strategies for a point mobile automaton moving amidst unknown obstacles of arbitrary shape
- Lumelsky, Stepanov
- 1987
(Show Context)
Citation Context ...m developed by Kamon, Rimon, and Rivlin [23] in 1995, whose name in turn acknowledges the globally-convergent, reactive nature of the “Bug” algorithms first developed by Lumelsky and Stepanov in 1=-=987 [40]-=-. These algorithms, and their exact relation to Wedgebug, are discussed in more detail in Section 1.2. The related RoverBug algorithm, also developed in this thesis, is the implementation of the Wedge... |

103 | A complete navigation system for goal acquisition in unknown environments
- Stentz, Hebert
- 1995
(Show Context)
Citation Context ...n the area of rover navigation has concentrated on heuristic local trajectory generation, and particularly obstacle avoidance. This style of navigation has also been referred to as “local navigation=-=” [79] and “-=-piloting” [43], especially when paired with a goal-seeking behaviour. Examples include: • Gat’s implementation of his ATLANTIS architecture (1991) [12], utilises as10 heuristic planner based on ... |

94 |
A Mobile Automaton: An Application of Artificial Intelligence Techniques
- Nilsson
- 1969
(Show Context)
Citation Context ...irst two categories seek to create maps or channels for robot navigation (see Fig. 1.1(a)). Examples of roadmap algorithms include the visibility and reduced visibility (also known as tangent) graphs =-=[57],[-=-36], Voronoi graphs [58], and Canny’s silhouette method [7]. Each of these schemes constructs a set of one-dimensional curves which encapsulate the topology of the free space, and serve as a system ... |

92 |
A Retraction Method for Planning the Motion of a Disc
- O’Dunlaing, Yap
- 1982
(Show Context)
Citation Context ...o create maps or channels for robot navigation (see Fig. 1.1(a)). Examples of roadmap algorithms include the visibility and reduced visibility (also known as tangent) graphs [57],[36], Voronoi graphs =-=[58], and -=-Canny’s silhouette method [7]. Each of these schemes constructs a set of one-dimensional curves which encapsulate the topology of the free space, and serve as a system of “freeways,” complete wi... |

84 | Obstacle detection for unmanned ground vehicles: a progress report", Robotics Research
- Matthies, Kelly, et al.
- 1996
(Show Context)
Citation Context ...which produced range images � from full stereo. The algorithms were originally developed by Larry Matthies and coded by Todd Litwin; the versions used on-board Rocky7 were adapted by Bob Balaram [49=-=],[48]-=-. In brief, given a stereo image pair, the image processor measures image similarity via correlation windows within a fixed disparity search range, then estimates the disparity for each pixel independ... |

81 | A Combinatorial Introduction to Topology - Henle - 1979 |

78 | Sensor based motion planning: The hierarchical generalized voronoi graph
- Choset
- 1996
(Show Context)
Citation Context ...tchy information may exist—are greatly facilitated by robots which are able to not only sense their environs, but to react accordingly as they navigate toward their goal. Sensor-based motion plannin=-=g [9], as-=- a field, arose to address this issue. The basic sensor-based “find-goal” problem is formulated as follows: the robot is situateds2 in unknown terrain, usually with an associated coordinate frame.... |

75 | Robot Navigation in Unknown Terrains: Introductory Survey of Non-Heuristic Algorithms
- Rao, Kareti, et al.
- 1993
(Show Context)
Citation Context ...d world model incrementally, such as Stentz’ D* algorithm [78], [77]. The third approach harks more closely to the heuristic planners, and includes the “Bug” algorithms of Lumelsky and Stepanov =-=[40], [64]-=-, which combine reactive behaviours with global parameters to reach the goal. All of these methods maintain provable properties of correctness and completeness, yet are fully applicable to unknown ter... |

66 |
Real Analysis, third edition
- Royden
- 1988
(Show Context)
Citation Context ...that the boundaries are continuous and rectifiable.) We assume that the reader has basic knowledge of real analysis, differential geometry, and of topology; good references for these topics are [69], =-=[68], [2-=-9] (for real analysis), [31] ∗ or rather basins of attraction, as will be shown later † Specifically, Kamon, Rimon, and Rivlin state that there are a finite number of local minima in the (assumed ... |

64 | Mars microrover navigation: Performance evaluation and enchanement. Autonomous Robots, 1995-12-01. Volume 2 Issue 4
- Matthies, Gat, et al.
(Show Context)
Citation Context ...rm of automatic gaze control: if no forward path can be found, the rover will sense a region to the left and right of the current visible patch. • the Go-to-Waypoint algorithm developed at JPL (1995=-=) [47]-=-, [82], [54] and implemented on the Rocky3.2 and Rocky7 prototype microrovers, as well as on the Sojourner rover now on Mars. This algorithm cycles through a list of behaviours (see Table 1.1) in orde... |

62 | A behavior-based system for off-road navigation
- Langer, Rosenblatt, et al.
- 1994
(Show Context)
Citation Context ...t, continuous driving along a predetermined path in fair terrain; a situation not applicable to the upcoming Mars microrover missions. • Langer, Rosenblatt, and Hebert’s behaviour-based system (19=-=94) [35]-=-, using Ganesha to manage the local map grid and to detect untraversable cells, and DAMN to arbitrate between votes on steering directions contributed by a set of behaviours. For off-road navigation, ... |

56 | Stochastic performance modeling and evaluation of obstacle detectability with imaging range sensors - Matthies, Grandjean |

54 |
Reliable Goal-Directed Reactive Control of Autonomous Mobile Robots
- Gat
- 1991
(Show Context)
Citation Context ...lso been referred to as “local navigation” [79] and “piloting” [43], especially when paired with a goal-seeking behaviour. Examples include: • Gat’s implementation of his ATLANTIS architec=-=ture (1991) [12], utilis-=-es as10 heuristic planner based on Slack’s “navigation templates” [75] for obstacle avoidance and goal seeking, as well as a version of Barraquand and Latombe’s non-holonomic planner [4] to re... |

52 | Optimal and efficient path planning for unknown and dynamic environments
- Stentz
- 1995
(Show Context)
Citation Context ...vigator, SMARTY (similar to the Langer, Rosenblatt, and Hebert navigator described above), and the DAMN voting arbitrator [79]. D* has been proven to be resolution-complete † in a bounded environmen=-=t [76]. • -=-Chatila and Lacroix’ 2D navigation algorithm (1995) [8], which has been implemented on the Adam prototype rover. From the sensed information, the † That is, D* is complete, assuming the resolution... |

38 | and Yu-Rwei Gwo. Dynamic motion planning of autonomous vehicles
- Shiller
- 1991
(Show Context)
Citation Context ...simulated static placements of a given rover model on the terrain at each node. These methods assume full prior knowledge of the terrain.s14 • Shiller, Gwo, and Chen’s B-spline optimal planner (19=-=90) [72]-=-, which models the terrain as a B-patch, generates the best K initial paths by graph search using kinematic constraints, then filters these paths using dynamic constraints and proximity to another pat... |

32 | Framed-quadtree path planning for mobile robots operating in sparse environments
- Yahja, Stentz, et al.
- 1998
(Show Context)
Citation Context ... method developed by Stentz (1993) [77], [78]. This method (described in Section 1.2.1) has been implemented on a HMMWV and on a prototype rover. It has recently been extended by Yahja, et al. (1998) =-=[89]-=- to a framedquadtree version, in order to reduce the memory requirements of the world model and to solve the inability of D* to plan some straight diagonal paths. In addition, the D* planner has been ... |

31 | B.Dacre-Wright, A practical motion planner for all-terrain mobile robots
- Simeon
- 1993
(Show Context)
Citation Context ...ing literature have not been implemented to the author’s knowledge, but have been demonstrated in simulation: • the similar extended elevation grid methods developed by Simeon and Dacre Wright (19=-=92) [73]-=- and by Kubota et al. (1995,7) [33], [34]. Both of these methods use a grid search method to select a path, based upon simulated static placements of a given rover model on the terrain at each node. T... |

30 | Critical Point Theory in Global Analysis and DifJkrential Topology - Morse, Cairns - 1969 |

28 |
Traversability analysis and path planning for a planetary rover. Autonomous Robots 6(2
- Gennery
- 1999
(Show Context)
Citation Context ...path. As a final step, perception tasks (for localisation and further terrain classification) are planned along this path. • the breadth-first, parallel grid search method developed by Gennery (1998=-=) [13]-=-. This algorithm uses a parallel search method derived from Witkowski [88] on an extended elevation grid which includes information on slope and terrain roughness at each cell. At the end of the searc... |

26 |
Commande Dynamique dans l’Espace Opérationnel des Robots Manipulateurs en Présence d’Obstacles. PhD thesis, l’École Nationale Supérieure de l’Aéronautique et de l’Espace
- Khatib
- 1980
(Show Context)
Citation Context ...—the original philosophy still held that the entire world model was known. In addition, though Khatib’s original formulation of the potential field method did not possess the property of completen=-=ess [27]-=-, further development of the theory has produced a classical potential field algorithm which is complete [66]. As previously stated, the classical planners as a group possess the key advantages of pro... |

26 |
Experiences with Operations and Autonomy of the Mars Pathfinder Microrover
- Mishkin, Morrison, et al.
- 1998
(Show Context)
Citation Context ...heuristic planners, such as Brooks’ subsumption architecture [5] or the track arbitration schemes developed at CMU [26],[78], as well as the “Go To Waypoint” algorithm employed by the Sojourner =-=rover [54]-=- and the Rocky7 rover [82] (see Table 1.1), share the useful property of being able to be made sensorbased much more easily than the classical planners, and can be applied to unknown terrains. These p... |

24 |
Shortest paths in the plane with convex polygonal obstacles
- Rohnert
- 1986
(Show Context)
Citation Context ...d of x, the q goalsT O1 O7 robot O O 5 2 O6 O4 T g O3 47 Figure 3.2). It can be proven that the shortest possible path that connects qinit to qgoal through F must be included within the Tangent Graph =-=[67]-=-. So, a planner need only search this graph to find the globally optimal path (measured by path length). However, if the planner does not already have full prior knowledge of its environment, the robo... |

24 |
The rocky 7 mars rover prototype
- Volpe, Balaram, et al.
(Show Context)
Citation Context ...s Brooks’ subsumption architecture [5] or the track arbitration schemes developed at CMU [26],[78], as well as the “Go To Waypoint” algorithm employed by the Sojourner rover [54] and the Rocky7 =-=rover [82]-=- (see Table 1.1), share the useful property of being able to be made sensorbased much more easily than the classical planners, and can be applied to unknown terrains. These planners dispense with the ... |

20 |
Construction of c-space roadmaps from local sensory data. What should the sensors look for? Algorithmica 17(4):357–379
- Rimon
- 1997
(Show Context)
Citation Context ...assical methods to a local sensed region. One set of methods incrementally builds “roadmaps” within the free space in the visible area, such as Choset’s HGVG [9] and Rimon’s adaptation of Cann=-=y’s OPP [65]. Of-=- note is the “Tangent Bug” algorithm, developed by Kamon, Rivlin, and Rimon [23]. The second approach springs from approximate cell decomposition, filling in a grid-based world model incrementally... |

19 | A New Range-Sensor Based Globally Convergent Navigation Algorithm for Mobile Robots
- Kamon, Rimon, et al.
- 1995
(Show Context)
Citation Context ...butes a proof of a key result required for the verification of completeness for the Wedgebug algorithm (and for the TangentBug algorithm which inspired this approach). The original proof, provided in =-=[21]-=- by Kamon, Rimon, and Rivlin, was slightly flawed, and further, was restricted to bounded workspaces populated by a finite number of obstacles, each with a smooth boundary. In this thesis, we not only... |

17 | An Approach to Rough Terrain Autonomous Mobility
- Kelly, Stentz
- 1998
(Show Context)
Citation Context ...mplemented on a HMMWV, and shares the ACCN system’s susceptibility to dead-ends (i.e., the algorithm is not complete). • RANGER, developed in Singh and Kelly (1995) [74] and Kelly and Stentz (1997=-=,8) [24], -=-[25]. This system, implemented on both a HMMWV and CMU’s prototype lunar rover RATLER, also arbitrates between inputs from hazard avoidance and goal-seeking behaviours. It optimises the behaviour of... |

16 |
Situationally Driven Local Navigation for Mobile Robots
- Slack
- 1990
(Show Context)
Citation Context ...y when paired with a goal-seeking behaviour. Examples include: • Gat’s implementation of his ATLANTIS architecture (1991) [12], utilises as10 heuristic planner based on Slack’s “navigation tem=-=plates” [75] for-=- obstacle avoidance and goal seeking, as well as a version of Barraquand and Latombe’s non-holonomic planner [4] to recover from cul-de-sacs. The planner was implemented on JPL’s Robby, a large pr... |

15 |
Path planning using a tangent graph for mobile robots among polygonal and curved obstacles
- Liu, Arimoto
- 1992
(Show Context)
Citation Context ... a line, that line is tangent § to obstacles containing the segment’s endpoints (see Figure 3.2: The Tangent Graph. † The Tangent Graph can also be constructed for curved 2D obstacles, as detaile=-=d in [38]. ‡ i.e., a vertex-=- v of an obstacle O such that in a neighborhood N ∈ O of v, the segment {x|x = λx1 +(1−λ)x2;0≤λ≤1}∈O, ∀x1,x2 ∈N. § A line l is tangent to a C-obstacle CB at a vertex x iff in a neigh... |

14 |
Fast Cross-country Navigation on Fair Terrains
- Pagnot, Grandjean
- 1995
(Show Context)
Citation Context ...d of the vehicle require significant computational resources, not available on the current batch of flightlike Mars microrovers. • Pagnot and Grandjean’s fast cross-country navigation approach (19=-=95) [60], whic-=-h computes a (typically trapezoidal) “trace” of possible rover trajectories (including uncertainty in sensing, motion control, and localisation) in each of 20 directions, incremented by 1 ◦ step... |

13 |
Size-frequency Distributions of Rocks on Mars and Earth Analog Sites: Implications for Future Landed Missions
- Golombek, Rapp
- 1997
(Show Context)
Citation Context ...ourtesy JPL) †† 3% of the ground in the MPF site was covered by rocks >20 cm tall, on average. The maximum concentration was 20%, in the “Rock Garden” [14],[16]. The predicted coverage was rou=-=ghly 8% [15]; fo-=-r comparison, VL1 had 4% coverage, and VL2 11% [85], [86]. ‡‡ The batteries, used during the mission primarily for nighttime operation and for ondemand power boosts during traverses, lasted 56 sol... |

12 |
Principles of Mathematical Analysis, second edition
- Rudin
- 1964
(Show Context)
Citation Context ...ssume that the boundaries are continuous and rectifiable.) We assume that the reader has basic knowledge of real analysis, differential geometry, and of topology; good references for these topics are =-=[69], [6-=-8], [29] (for real analysis), [31] ∗ or rather basins of attraction, as will be shown later † Specifically, Kamon, Rimon, and Rivlin state that there are a finite number of local minima in the (as... |

11 |
Rough terrain autonomous mobility—part 2: An active vision, predictive control approach
- Kelly, Stentz
- 1998
(Show Context)
Citation Context ...nted on a HMMWV, and shares the ACCN system’s susceptibility to dead-ends (i.e., the algorithm is not complete). • RANGER, developed in Singh and Kelly (1995) [74] and Kelly and Stentz (1997,8) [2=-=4], [25]. -=-This system, implemented on both a HMMWV and CMU’s prototype lunar rover RATLER, also arbitrates between inputs from hazard avoidance and goal-seeking behaviours. It optimises the behaviour of follo... |

11 | Robot planning in the space of feasible actions: Two examples
- Singh, Kelly
- 1996
(Show Context)
Citation Context ...radius. This system has also been implemented on a HMMWV, and shares the ACCN system’s susceptibility to dead-ends (i.e., the algorithm is not complete). • RANGER, developed in Singh and Kelly (19=-=95) [74] a-=-nd Kelly and Stentz (1997,8) [24], [25]. This system, implemented on both a HMMWV and CMU’s prototype lunar rover RATLER, also arbitrates between inputs from hazard avoidance and goal-seeking behavi... |

8 |
T.Yoshimitsu, Path planning for planetary rover based on traversability probability
- Kubota
- 1995
(Show Context)
Citation Context ...ented to the author’s knowledge, but have been demonstrated in simulation: • the similar extended elevation grid methods developed by Simeon and Dacre Wright (1992) [73] and by Kubota et al. (1995=-=,7) [33]-=-, [34]. Both of these methods use a grid search method to select a path, based upon simulated static placements of a given rover model on the terrain at each node. These methods assume full prior know... |

7 |
Sojourner on mars and lessons learned for future planetary rovers
- Wilcox, Nguyen
- 1998
(Show Context)
Citation Context ...ary drivers for memory are the same as for CPU choice, plus cost (e.g., a single shielded PROM costs $30K) and volume [84].s30 Figure 2.12: Map of Sojourner’s traverse, from rover odometry, Sols 1-7=-=8 [87]. (c-=-ourtesy Rick Welch) flight-qualified electronics, will continue to constrain the computing capabilities— and therefore the complexity of on-board autonomy—of future rovers. Of note, in order to ma... |

7 |
A Parallel Processor Algorithms for Robot Route Planning
- Witkowski
- 1983
(Show Context)
Citation Context ...ain classification) are planned along this path. • the breadth-first, parallel grid search method developed by Gennery (1998) [13]. This algorithm uses a parallel search method derived from Witkowsk=-=i [88]-=- on an extended elevation grid which includes information on slope and terrain roughness at each cell. At the end of the search, the total cost of the optimum path through each node in the local grid ... |

6 |
Passive stereo range imaging for semi-autonomous land navigation
- Matthies
- 1992
(Show Context)
Citation Context ...edge, with the rover at the apex. Although the mast can certainly image regions close to the rover, the limitations of the Rocky7 implementation of the obstacle detection algorithm used (described in =-=[45], [48]-=-) do not allow for robust obstacle detection within roughly 1.5 m of the rover. A choice must be made whether to use a single mast “wedge view” for path planning, or whether to let a single “wed... |