Background
-
Mobility in unstructured environment The robot should traverse complex geometric ground, such as rough terrain, unstable and unstructured ground, high step, and obstacles. To end this, the robot must have a suitable physical structure and a control system to provide stability and driving force to arbitrary directions, in particular, vertical upward.
-
Mobility in unknown (low-visibility) environment The robot should move even when the visibility for human operators and/or robot itself is quite low. In extreme environment with heavy smoke and fog, the vision sensors for localization, mapping, and navigation cannot obtain sufficient environmental information. Moreover, even when such vision sensors are broken, the robot keeps to move for continuing tasks or returning to the rescue base.
Classification of locomotion modes
Crawler-crawling mode (CCM)
Arm-walking mode (AWM)
Compound-locomotion mode (CLM)
The advantages of CLM
Better mobility
Wider use range
More stable locomotion state
Requirements of CLM
Requirements of control system
-
Control system should have the capability to understand key parameters of surrounding environment, such as obstacle height and the distance between robot and obstacle.
-
Control system could monitor robot stable state in real time, and maintain the robot state to be stable by controlling joints appropriately. This is because work site may be irregular and slipping the endpoints of arms is unavoidable when contacting it to the environment.
-
Control system could monitor robot stability margin real time to avoid accidents.
Preparation: coordinate systems
i
|
a
i–1
|
\(\partial_{i - 1}\)
|
d
i
|
θ
i
|
---|---|---|---|---|
1 | 0 | 0 | 0 |
θ
1
|
2 | 90° | 0 | 0 |
θ
2
|
3 | 0 | 490 (l1) | 78 (d3) |
θ
3
|
Terrain exploration (estimation of height of obstacle)
-
Case 1: Flat terrain\(\left| {{z}_{1} - {z}_{4} } \right| <{\sigma}_{1}\) and \(\left| {{z}_{2} - {z}_{3} } \right| < { \sigma }_{1}\).Robot moves on a flat landscape. And the height different between four EPRAs is small than σ1.
-
Case 2: Upward step or slope\({\sigma}_{1} < \left( {{z}_{1} - {z}_{4} } \right) < { \sigma }_{2}\) and \({\sigma}_{1} < \left( {{z}_{2} - {z}_{3} } \right) < { \sigma }_{2}\).There is an upward step or obstacle in front of the robot. According to robot structure and driving force, robot has the ability to get over steps which height is lower than \(\sigma_{2}\).
-
Case 3: Downward step or pit\(-{\sigma}_{3} < \left( {{z}_{1} - {z}_{4} } \right) < -{ \sigma }_{1}\) and \(-{\sigma}_{3} < \left( {{z}_{2} - {z}_{3} } \right) < -{ \sigma }_{1}\)There is a downward step or a pit in front of robot. Robot has the ability to down steps which height is lower than \(\sigma_{3}\).
Recognition of COG position of robot
Control system design for CLM
Crawler controls
-
Case 1: \(\left| {T^{\prime}} \right| < T_{0}\) (low torque)Control rule: \(A = A^{\prime}\)Measured torque will change in a certain range even if robot is running on a flat road, due to the vibration of robot. When robot is in this state, front flippers angle remain the same.
-
Case 2: \(\left| {T^{\prime}} \right| > T_{0}\) (high torque)Control rule: \(A = A^{\prime} - \Delta A*(T^{\prime}/\left| {T^{\prime}} \right|)\)ΔA is adjusting amplitude in each sample period. This parameter can be set according to the system sampling frequency and response time of flipper. When the robot is in this state, flippers will rotate depending on terrain state. For example, when robot encounters an obstacle, front flippers rotate up, while robot meets a pit, front flipper will rotate down.
-
Case 3: \(A^{\prime} > A_{0}\) (flippers in limitation position)Control rule: \(A = A^{\prime} - \Delta B\)A0 is front flipper limit value and ΔB is another adjusting amplitude for getting over obstacle. When robot is in this state, control system will consider there is an obstacle in front of robot. Combined with arms state, crawlers and arms will be controlled to get over obstacle, and this part will be detailed in later section.
Arms and crawlers control in rough terrain passing
EPRA | Initialization coordinate (when robot start) | Target coordinates (after arm transfer) |
---|---|---|
P1 | (540, 399, − 433) | (618, 399, − 433) |
P2 | (298, − 399, − 433) | (618, − 399, − 433) |
P3 | (− 419, 399, − 433) | (− 121, 399, − 433) |
P4 | (− 618, − 399, − 433) | (− 121, 399, − 433) |
-
Case 1: \(\left| {{P}_{{{i}\_{a}}} \left( {x} \right) - {P}_{{{i}\_{t}}} ({x})} \right| > \Delta {C}\)This situation means that there are obstacles between the arm and target position.
-
Case 2: \(\left| {{P}_{{{i}\_{a}}} \left( {x} \right) - {P}_{{{i}\_{t}}} ({x})} \right| < \Delta {C}\) and \({P}_{{{i}\_{a}}} \left( {z} \right) - {P}_{{{i}\_{t}}} ({z}) > \Delta {D }\)This means there are obstacles on the target position, and the EPRA stopped on the top of obstacle.
-
Case 3: \(\left| {{P}_{{{i}\_{a}}} \left( {x} \right) - {P}_{{{i}\_{t}}} ({x})} \right| < \Delta {C }\) and \({P}_{{{i}\_{a}}} \left( {z} \right) - {P}_{{{i}\_{t}}} \left( {z} \right) < - \Delta {D }\)This means that there is a pit in the target position, and arm endpoint stopped at the bottom of this pit. In the last phase of transferring arm, EPRA will keep moving downward until control system detects stop signal or the arm joints reach the limit position. If arm still does not contact with ground when joints reach the limit position, it means where there is a deep pit. In this case, a new target point will be selected.
Arms and crawlers control in getting over obstacle
Experimental setting
Control system and hardware
System | Item | Specification |
---|---|---|
Total | Length × width × height (mm) | 1800 × 1100 × 1700 |
Weight (kg) | 100 | |
Degree of freedom | 30 | |
Arm | Number of arms | 4 |
Degree of freedom per one arm | 6 (within 1 grab) | |
Crawler | Number of crawlers | 6 (within 4 flippers) |
Degree of freedom | 6 (flippers: active) | |
Maximum speed (mm/s) | 63 | |
Climbing ability (°) | 30 | |
Sensor | Encoder | 22 |
Inertia measurement unit (IMU) | 1 | |
Wireless CCD camera | 1 | |
Operating | Number of control lever (channel) | 4 (7 DOF) |
Interface | Control pedal | 8 (4 spare) |
Communication method | TCP/IP |
Tasks setting
-
Getting over obstacle (test for mobility in unstructured environment) To facilitate comparison and analysis, we simplified the obstacle into step. The robot that can get over higher step has better obstacle get over capability. For testing CLM control mode, in this experiment, robot needed to finish three fundamental tasks in disaster response works, respectively, climbing a one-terrace step, climbing a two-terrace step from the front and climbing a two-terrace step from the side. According to the structure of OCTOPS, as shown in Fig. 3, OCTOPUS can climb the step which height is lower than 250 mm (the length of flippers) without using arms. Thus, we set the height of those steps are 400 mm [8], which is more than twice as steps in residential areas. Two locomotion modes (CCM and CLM) were used to get over one-terrace step, and CLM mode also be used to execute the other two tasks introduced above. For CCM, robot is manually controlled by two operators according to the sufficient feedback visual images. But for CLM, just one operator input move direction based on an image got form in-vehicle camera (the image shows the environment in front of robot), and CLM control system knows nothing about the environment in advance.
-
Passing through rough road (test for mobility in unknown environment) In this experiment, a 500 mm height concrete block was in the way of robot moving forward. The distance between robot right side and concrete block was about 10 mm. Some robot arms would collide with this obstacle and automatically adapt it. These information is also unknown by control system.