Surmounting obstacles by arm maneuver for unmanned power shovel
 Peshala G. Jayasekara^{1}Email author and
 Hitoshi Arisumi^{1}
DOI: 10.1186/s4064801500387
© Jayasekara and Arisumi. 2015
Received: 30 June 2015
Accepted: 13 October 2015
Published: 29 October 2015
Abstract
Large debris created after natural disasters restrict access to inner parts of affected regions, and slows down disaster relief operations. Power shovels are often used to clear wreckage but the process can take a lot of time. Moreover, it is dangerous to involve human workers operating heavy machinery in such unstable conditions. To speedup access to inner areas, obstacles can be surmounted with the assistance of carefully maneuvered power shovel arm, instead of removing them. In this work, an autonomous obstacle surmounting technique for an unmanned power shovel is proposed. Out of different sequences, the one that optimizes the total energy consumption is chosen as the best candidate for surmounting a given steplike obstacle. Dynamic simulation results show the effectiveness of the proposed method.
Keywords
Power shovel Obstacle surmounting Unmanned OptimizationBackground
Each year many countries in the world are challenged by natural disasters such as earthquakes, floods, tsunamis, typhoons and so on. While these disasters can cause loss of life, they also generate large amounts of debris, which further reduces access to inner parts of the affected regions. This results in slowing down of search and rescue missions as well as other disaster relief operations. Power shovels are used in disaster stricken areas to remove wreckage and to clear up roads. Nevertheless, it is a dangerous task to involve human workers to operate such heavy machines in these conditions as there is chance of tip over of these machines due to instability. Moreover, it can take a lot of time to clear up all the obstacles and finally get access to inner areas.
In order to speed up disaster relief operations, large obstacles can be surmounted instead of clearing them. Using crawler wheels, power shovels have the ability to go over objects effectively compared to other vehicles. However, there exists several challenges to obstacle surmounting with power shovels as follows: only the obstacles that are within the climbing limit of the crawler can be overcome; crawler power can be easily saturated; power shovel can tip over and damage itself if not moved skillfully when it is on top of an obstacle; being heavy machines, power shovels are also difficult to get back to the original state at the event of tip over.
Previous works on excavators can be found at [2–5]. Ref. [2] presents a dynamic model for an excavator with the intention of developing an automated excavation control system for terrestrial, lunar, and planetary excavation. The kinematics of excavators having hydraulic actuators are investigated in detail in [3]. Ref. [4] mainly focuses on a system that completely automates the truck loading task. In that system the excavator’s software decides where to dig in the soil, where to dump in the truck, and how to quickly move between these points while detecting and stopping for obstacles. The nature of an excavation process and the way it may be controlled is investigated with the intention of automatic excavation in [5].
Among other related works, an autonomous staircase climbing tracked mobile robot is introduced in [6]. However, it does not use any arms for assistance. Ref. [7] introduces a pilot system for a rescue robot to let a human operator suggest good directions to traverse on a 3D debris environment. In contrast, the proposed system seeks autonomous operation without any human intervention. Ref. [8] has proposed a wheeled robot with a movable center of mass (CoM) to ease the traverse over rough terrain. The proposed system also considers change of CoM, indirectly, by maneuvering its arm; however, the main focus is on getting the reaction force at the bucket as a support to lift the power shovel crawler.
Problem statement
Given a steplike obstacle, the power shovel should maneuver its arm to climb up the obstacle in a smooth trajectory. Also, it should carry out the maneuver in such a way that the total energy consumption is minimized. The whole process is to be automated so that the risk for the human workers is eliminated.
Nomenclature
The basic parts of a power shovel are illustrated in Fig. 1; boom, arm and bucket are the main links considered, while each link is attached to the previous link by boomjoint, armjoint and bucketjoint, respectively. These joints are individually controlled to achieve different poses similar to a serial link robot manipulator. There is a rotating platform, which can rotate around its local z axis (vertical) so that the whole boomarmbucket composite can be moved to its front and back, symmetrically. The vehicle stands on a crawler that provides better mobility on unstructured terrain.
Methods
Trajectory
Considering the initial and final poses in each subfigure of Fig. 2, a suitable trajectory should be generated for the motion of the power shovel. The 1st stage is materialized using a single trajectory (from here onwards, mentioned as sequence 1) while the 2nd stage is further divided into two trajectories (sequence 2–1 and sequence 2–2). To formulate smooth trajectories, the 3D motion of the bottom center position of the crawler \(P_b\) is considered.
Sequence 1
Sequence 2–1
Sequence 2–2
Joint angle calculation
In order to achieve the required crawler trajectories, the power shovel joints should be controlled appropriately. In contrast to a conventional serial link manipulator operation, it is interesting to note that in this maneuver, the end effector (bucket) remains stationary while the base (crawler) is moving. Furthermore, it can be shown that inverse kinematics can be calculated in closed form. This section describes the procedure for calculating the reference joint angles for each joint.
Considering the power shovel local coordinates, let the boom joint position be \(p_1=[x_1\;z_1]^T\), the bucket joint position be \(p_2=[x_2\;z_2]^T\) and the arm joint position be \(p_3=[x_3\;z_3]^T\) (y coordinate is assumed to be constant throughout the whole maneuver). In both stages, \(p_1\) and \(p_3\) points change with time, while \(p_2\) remains stationary. As given in Fig. 6, the joint angles \(\theta _1\), \(\theta _2\) and \(\theta _3\) represent boom, arm and bucket joint angles, respectively; \(\theta ^*_1\) and \(\theta ^*_2\) serve as auxiliary angles in the calculation process.
Crawler wheel reference speeds
To retain the required trajectory, the ground touching crawler wheels need to maintain certain speed based on the sequence in operation. The reference angular speed of the crawler wheel can be calculated in two steps. First, the velocity of the wheel center position, \(P_c\) (Fig. 5), is calculated based on \(P_b\) position and taking the time derivative of it. Next, by assuming that the wheel undergoes no slippage and maintains perfect rolling, the reference angular speed is calculated using the wheel radius.
Sequence 1
Sequence 2–1
In order to pitch about the edge of the steplike structure, the crawler wheels should be locked in this sequence. Therefore, all the wheel reference speeds are set to zero.
Sequence 2–2
Optimization
Several parameters that can affect the nature of the total obstacle surmount operation can be identified. Not only the power shovel trajectories in different sequences, but also the total energy consumption change depending on these parameters. The upcoming sections provide more details on the parameters and how they can be optimized for optimal energy requirements.
Parameters
As illustrated in Fig. 2, the parameters are as follows: P1 is the distance between the initial power shovel bottom center (\(P_b\)) and the edge of the steplike structure (E) in the 1st stage; P2 is the distance between the bucket position (B) and E in the 1st stage; P3 is the distance between B and E in the 2nd stage; P4 is the distance between the final \(P_b\) position and E in the 2nd stage; \(\alpha _F\) is the lean angle of the power shovel after the 1st stage. The height of the steplike structure is treated as a known constant, which has been set to the length of the diameter of the crawler wheel, plus the thickness of the crawler plate. In a real scenario, this step height can be obtained using an exteroceptive sensor such as a stereo camera or a LIDAR system. To obtain optimal parameters that minimize the total energy consumption in the complete obstacle surmount operation, sequential quadratic programming (SQP) technique has been used. The parameter P4 was also treated as a constant in order to minimize the computational overhead.
Objective function
Constraints

Maximum stretch This linear constraint imposes a maximum length the power shovel can stretch its arm, where D is a power shovel specific constant.Equation (21) is obtained by adding \(P1 + P2 \le D\) and \(P3 + P4 \le D\) inequalities for the two stages.$$\begin{aligned} P1 + P2 + P3 + P4 \le 2D \end{aligned}$$(21)

Singular configurations Once the bucket position (P2 or P3) is decided or fixed, the power shovel should have a valid joint configuration at any given time instance in the trajectory to reach it. Based on intersection of circles having radii equal to effective boom length (Fig. 6) and arm length, the following two nonlinear constraints can be imposed:$$\begin{aligned} (x_1x_2)^2+(z_1z_2)^2\le (L_1+L_2)^2 \end{aligned}$$(22)where \(x_1\), \(x_2\), \(z_1\), \(z_2\), \(L_1\) and \(L_2\) are described as in Joint Angle Calculation section. \(\left. x_i\right _{i=1,2}\) and \(\left. z_i\right _{i=1,2}\) are derived from \(\left. Pi\right _{i=1,2,3,4}\) and \(\alpha _F\).$$\begin{aligned} (x_1x_2)^2+(z_1z_2)^2\ge (L_1L_2)^2 \end{aligned}$$(23)

Collision The power shovel must not collide with the steplike structure throughout the obstacle surmount operation. Since the end effector (bucket) is fixed, collision is checked between the steplike structure and the crawler of the power shovel.
Results
Specifications: miniature power shovel
General  Specific  Value 

Mass  Crawler mass  4.24 kg 
Platform mass  4.53 kg  
Boom mass  0.9 kg  
Arm mass  0.29 kg  
Bucket mass  0.47 kg  
Length  Boom effective length  0.359 m 
Arm length  0.171 m  
Bucket length  0.11 m  
Crawler wheel radius  0.033 m  
Step height  0.066 m 
Parameters: ODE
Parameter  Value 

Time step  0.002 s 
Friction coefficient  0.7 
Algorithm  Dantzig’s (dWorldStep) 
Constraint force mixing (cfm)  1e−8 
Error reduction (erp)  0.2 
Exhaustive fast batch simulations
Starting from the lower bounds, first, an exhaustive batch simulation was carried out using the following parameter increment sizes: P1_increment_size = P2_increment_size = P3_increment_size = 0.03 m, \(\alpha _F\)_increment_size = 3.0° to obtain an initial point for the SQP optimization phase. These relatively large increment sizes help to reduce the time taken for this batch simulation. The initial point for the SQP optimization turned out to be P1 = 0.2675 m, P2 = 0.21 m, P3 = 0.4396 m, and \(\alpha _F\) = 16.75° with the energy consumption of 18.658 J.
SQP optimization
Parameters: CFSQP Optimization
Parameter  Value 

Free variables  4 
Objective functions  1 
Iterations max.  500 
Final norm (\(\epsilon\))  1.0e−8 
Perturbation size (\(\lambda\))  0.01 
Exhaustive slow batch simulations
To verify the results obtained in the SQP optimization phase, another exhaustive batch simulation was carried out, this time using smaller parameter increment sizes: P1_increment_size = P2_increment_size = P3_increment_size = 0.01 m, \(\alpha _F\)_increment_size = 1.0°. For the given specifications and for the given steplike structure, it was identified at the end of this heavy time consuming batch simulation that the obstacle can be surmounted with least energy consumption of 17.155 J with P1 = 0.2475 m, P2 = 0.22 m, P3 = 0.3996 m, and \(\alpha _F\) = 16.75° parameters.
Discussion
It can be observed that the CFSQP result compare well with the result obtained in the exhaustive slow batch simulation. In fact, CFSQP had captured the continuous nature of the objective function (energy) and as a result had been able to obtain the most energy efficient parameters, which cannot be seized even when exhaustive batch simulations are run based on smaller parameter increments. This is evident from CFSQP energy result being smaller than that of the exhaustive slow batch simulations.
To make CFSQP running, a modification had to be included into the original algorithm. That is, when the gradient of the objective function is calculated by finite difference approximation, the algorithm tends to use nearby points, which can even violate the given constraints. This can make ODE crash due to illdefined inputs. To overcome this problem, a large cost was injected when illdefined points were tested.
The parameter P4 is considered to be constant in the simulation. The only requirement for P4 is such that the power shovel should be able to maintain stability once it lifts the bucket after sequence 2–2 is completed. Therefore, as long as this requirement is fulfilled, P4 can be set to an empirical value and not be included in the optimization process.
The open dynamics engine (ODE) [9] consists of two default solvers namely, (1) Dantzig’s Agorithm solver, and (2) Successive OverRelaxation (SOR) Projected GaussSeidel (PGS) LCP solver. Dantzig’s Agorithm solver has been used in this work as it attempts to achieve a numerically exact solution, even though it is about one order of magnitude slower than SOR PGS LCP solver.
As the first step of obstacle surmount operation a simple steplike structure was chosen as the obstacle. However, one should not forget that power shovels have the advantage of modifying the terrain using its arm and bucket. As a result, it can pave the unstructured terrain in front of it to shape it to a steplike structure using fast methods like bench cut method and employ the proposed method; totally modifying the terrain into a smooth slope would take a considerable amount of time, which negates the original purpose.
Conclusion
Summary
Power shovels serve as essential machinery to remove wreckage and help accessing inner parts of disaster stricken areas. However, instead of removing obstacles, they can be surmounted to save time. In this work, an arm maneuver based method is proposed to surmount steplike obstacles for unmanned power shovels. A smooth trajectory was formulated to minimize jerk caused by the power shovel motion. Using SQP optimization technique the most energy efficient motion profile was identified, which was confirmed by exhaustive batch simulations. The proposed method was verified by conducting 3D dynamic simulations and its effectiveness has been demonstrated. The authors believe that this is the first time a study has been conducted on unmanned power shovel for obstacle surmounting using arm maneuver mechanism.
Future work
The proposed method is the first step towards solving the obstacle surmount maneuver and therefore, a simple steplike obstacle was chosen. Due to the symmetry of the problem, the Ycoordinate motion could be overlooked in the simulations. Moreover, boom, arm and bucket joint axes are parallel to each other, which make the corresponding links to move in a plane. This also contributed to simplifying the 3D motion into the 2D domain. Debris created after natural disasters, however, have very complex shapes. The authors hope to address overcoming of different kinds of obstacles in the future, which will necessitate analysis of a full 3D motion. Another aspect to be investigated when overcoming complex obstacles is the stability issue. The power shovel should maintain its stability throughout the operation and should not tip over under any circumstances.
Simple higher order polynomials were employed in trajectory planning due to its simplicity and ease of use, and no problems were encountered. Nevertheless, splines (piecewise continuous polynomials) have gained popularity in trajectory planning and the authors would like to observe the improvements, if any, by changing the simple polynomials into spline functions in the future.
As the next step, the authors expect to carry out experiments using the real hardware in the future to confirm the results obtained through simulations.
Declarations
Authors' contributions
PGJ carried out the design, simulations and drafted the manuscript. HA conceived the study, participated in its design and helped with resolving errors. Both authors read and approved the final manuscript.
Acknowledgements
The authors are grateful to Dr. H. Adachi, Prof. N. Koyachi and Dr. S. Sarata for their contribution in designing the crawler base frame.
Competing interests
The authors declare that they have no competing interests.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 Javiereis’s Channel. http://www.youtube.com/watch?v=P8p8XkTg2M
 Vähä P, Skibniewski M (1993) Dynamic model of excavator. J Aerospace Eng 6(2):148–158View ArticleGoogle Scholar
 Koivo A (1994) Kinematics of excavators (backhoes) for transferring surface material. J Aerospace Eng 7(1):17–32View ArticleGoogle Scholar
 Stentz A, Bares J, Singh S, Rowe P (1999) A robotic excavator for autonomous truck loading. Auton Robots 7(2):175–186View ArticleGoogle Scholar
 Hemami A (1995) Fundamental analysis of automatic excavation. J Aerospace Eng 8(4):175–179MathSciNetView ArticleGoogle Scholar
 Mihankhah E, Kalantari A, Aboosaeedan E, Taghirad HD, Ali S, Moosavian A (2009) Autonomous staircase detection and stair climbing for a tracked mobile robot using fuzzy controller. In: IEEE International Conference on Robotics and Biomimetics, ROBIO 2008, IEEE. pp 1980–1985 Google Scholar
 Magid E et al (2011) Controlled balance losing in random step environment for path planning of a teleoperated crawlertype vehicle. J Field Robot 28(6):932–949View ArticleGoogle Scholar
 Nakamura S, Faragalli M, Mizukami N, Nakatani I, Kunii Y, Kubota T (2007) Wheeled robot with movable center of mass for traversing over rough terrain. IEEE/RSJ Int Conf Int Robots Syst IROS 2007:1228–1233Google Scholar
 Smith R et al (2005) Open dynamics engineGoogle Scholar
 Quigley M et al (2009) Ros: an opensource robot operating system. In ICRA Workshop on Open Source Software, vol. 3, p 5Google Scholar
 Lawrence C, Zhou J, Tits A. (1997) User’s guide for cfsqp version 2.5. University of MarylandGoogle Scholar