Professional Documents
Culture Documents
Complete Floor Coverage in Cleaning Robots Using Lateral Ultrasonic Sensors
Complete Floor Coverage in Cleaning Robots Using Lateral Ultrasonic Sensors
Technology Conference
Vail, CO, USA, 20-22 May 2003
I. INTRODUCTION
Proposed algorithms for cleaning or complete coverage [1,
2] have always straight-line paths while the cleaning
procedure used to start next to a wall (or similar object).
However, a mobile robot has always problems in the
development of a path following trajectory. An application
like floor cleaning where the complete path can be very large
and plenty of turns requires a low positioning error profile or
a procedure for positioning error correction. Several methods
have been proposed for path following and error correction
[X, X]. However, most of these methods can not be used if
the planned robot must be classified as domestic or low cost
because of the huge computational requirements.
In this work a procedure for straight-line path following
without explicit landmarks using a lateral reference distance
vector is enunciated and tested in a differential drive cleaning
robot with embedded ultrasonic sensors (fig. 1). The problem
of the wide beam width of the transducers is also studied and
a methodology is proposed for the adaptation of the initial
lateral distance vector to make it useful regardless of the
robot relative distance to the walls.
II. STRAIGHT-LINE PATH FOLLOWING
The main problem for a cleaning robot is the coverage
algorithm in unknown scenario. There are two main coverage
algorithms that can be used in cleaning applications: spiral
[3] and boustrophedon (the way of the ox) [4].
The spiral coverage starts with a wall following stage at a
close distance to the wall and when the robot detects that has
reached a previously explored area, the wall following stage
continues but the distance to the wall is increased
approximately the robot width. Every turn the distance to the
wall is increased the same way. This algorithm runs very well
if all the objects located in the trajectory lean against the
--
--
--
--
--
--
--
200
Fig. 2. Starting with the robot in parallel to the wall. The distance
measured is stored in the reference lateral vector.
--
--
--
201
199
202
200
--
--
--
201
199
202
200
--
--
--
--
401
399
402
400
402
401
200
Fig. 4. When the robot reaches an area where the reference lateral
vector is available it is used for parallel alignment.
-153
-149
-150
201
199
202
200
--
--
--
--
--
--
--
--
350
-153
-149
-150
201
199
202
200
399
397
401
400
751
749
752
750
755
753
403
402
550
1.2
2.5
1.4
1
0.8
0.6
2
1.5
1
0.4
0.5
0.2
20
60
120
distance (cm)
180
240
1 cell gap
3 cells gap
6 cells gap
10
20
30
40
distance to the far wall (cells)
50
60
WALL
3.5
A
WRITING
DESK
2.5
1.5
1
Trajectory A
LRV
0.5
object at 3 cells from the wall
object at 9 cells from the wall
0
10
20
30
40
distance to the wall (cells)
--
--
--
--
173
172
173
174
174
173
175
175
175
175
175
175
175
175
175
175
ML
--
--
--
--
173
172
173
174
174
173
LRV
Inf
-80
-82
Inf
173
172
173
174
Trajectory C
174
173
MLRV
50
60
Fig. 9. Obtained corner shadow for a plain object (width: 3 cells, high:
10 cells) located at different distances from the wall.
MLRV
--
--
--
--
--
--
--
441
441
441
ML
Inf
187
185
Inf
439
439
441
441
440
440
LRV
Inf
-80
-82
Inf
173
172
173
174
Trajectory E
174
173
--
454
452
--
--
--
--
708
708
708
448
435
438
449
695
696
698
698
698
699
IV. TESTS
MLRV
ML
V. CONCLUSIONS
A procedure for straight-line path following using
ultrasonic sensors has been proposed. The procedure has been
tested in a limited set of cases in a typical office environment.
During the tests a foreseeable measurement problem has been
detected in the presence of rounded objects and the sensors
placement problem has also been detected. The method has
been proven useful for robot alignment in unstructured
environments but requires more intensive tests to detect other
possible special cases and the expansion to two dimensions.
REFERENCES
[1]
[2]
[3]
[4]
[5]