1、 1 Agricultural Robotic Platform with Four Wheels Steering for Weed Detection Thomas Bak； Hans Jakobsen Department of Agricultural Engineering, Danish Institute of Agricultural Sciences, Schoottesvej. 17, DK-8700 Horsens, Denmark；e-mail of corresponding author: tbcontrol.auc.dk (Received 10 January
2、2003； accepted in revised form 14 October 2003； Published online 23 December 2003) A robotic platform for mapping of weed populations in fields was used to demonstrate intelligent concepts forautonomous vehicles in agriculture which may eventually result in a new sustainable model for developedagric
3、ulture. The software implements ahybrid deliberate software architecture that allows a hierarchical decomposition of the operation. The lowestlevel implements a reactive feedback control mechanism based on an extension of simple control for car-likevehicles to the four wheel case. The controller des
4、ign forces the front and rear of the vehicle to follow a predetermined path and allows the vehicle to maintain a fixed orientation relative to the path. The controllerrationale is outlined and results from experiments in the field are presented. 1. Introduction Advances in mechanical design capabili
5、ties, sensingtechnologies, electronics, and algorithms for planningand control have led to a possibility of realizing fieldoperations based on autonomous robotic platformsThe need for such systems is driven by increasingfinancial pressure on farmers combined with publicconcern about the environment
6、and workingconditions. Efficient deployment of autonomousrobotic platforms in the field will allow care andmanagement of crops in a very different way fromwhat is known today (Belasco et al., 2002； Cho et al,2002). Robotic platforms and implements maysense and manipulate the crop and its environment
7、 ina precise manner with minimal amount of materials andenergy making them potentially more efficient thantraditional machinery. This is likely to reduce theenvironmental impact while increasing precision andefficiency (Kondo &Ting, 1998； DeBaerdemaekretal, 2001). The result is a new sustainable mod
8、el fordeveloped agriculture. This paper presents an overview of the system andapproach. Section 2 provides a system description. Thisincludes a description of modular mechanical concepts well as the Techtronic 2 implementation of thesystem. Everything is tied together in hierarchicalhybrid software
9、architecture. In Section 3, the focus ison a specific mobility control strategy that extendssimple controllers to 4WS. The result is a system thatallows the vehicle to track a given path, whilemaintaining the front and rear implement bars on thepath. Results from experiments in the field aresummariz
10、ed in Section 4 and demonstrate the effectiveness of the proposed 4WS solution. Finally, conclusions are drawn and discuss further research. This paper concentrates on the engineering aspects of the researchand evaluation of the experimental system. 2. System description The robotic platform describ
11、ed here is meant todemonstrate novel sensing capabilities (Sgaard &Olsen, 2000) and semi-autonomous operation of arobotic platform for agriculture. The immediate agronomic aim of the project is to demonstrate efficientmeasurement of spatial and temporal crop and weedmeasurements. Given that the vari
12、ability in weeds ismeasured and mapped, inputs can be varied accordingto a defined strategy providing environmental andeconomic benefits. Studies show that 5080% of thecosts for herbicides can be saved when treating onlypatches where weeds actually grow (Green et al., 1997；Nordmeyer et al., 1997). F
13、undamental for the success ofsuch a system is the integration into farm managementsystems, e.g. job creation and path planning (Srensenet al., 2002). 2.1. Robotic platform The basis for the robotic platform is the mobilitycapability provided by the wheel module mechanismshown in Fig. 1. Each of the
14、four identical wheelmodules include a brushless electric motor for propulsion that provide direct drive without gearing.Motor, amplifier and microcontroller are all mountedin the wheel hub. Steering capability is achieved by aseparate steering motor mounted on top of thewheel module shaft to create
15、a two-degree-of-freedommechanism. The steering motor amplifier and thecontrol electronics are mounted next to the steeringmotor. The control electronics (wheel node) are basedon a commercial agricultural job computer and handlethe local position servo control for the steering andprovide torque contr
16、ol of the drive motors. The drivermotor electronics allow speed and current (torque)feedback while the steering servo system provide asteering angle feedback derived from motor shaftencoders. 3 2.2. Platform electronics This allows programs to be built automatically andsubsequently execute in near r
17、eal-time on the platformcomputer. The solution supports transmission controlprotocol/internet protocol (TCP/IP) sockets for remotecommunication with the running code which allowmonitoring and modification of parameters duringdevelopment. 2.3. System architecture The system architecture adopted is si
18、milar to thehybrid deliberate approach (Arkin, 1990) that is nowcommon in mobile robotics systems (Orebaack. &Christensen 2003). The three-layer architecture consistsof: (1) a reactive feedback control mechanism thathandles stabilization and tracking, (2) a plan-executionmechanismthatdealswithe.g.tr
19、ajectorygenerationandtask decomposition, and (3) a mechanism forperforming time-consuming deliberative computations andinteraction with human operators, i.e. job creation.The hierarchical structure is shown in Fig.4. 3Mobility control The motion of the robot can always be viewed as an instantaneous
20、rotation around a time varying point called the instantaneous centre of rotation (ICR).Hence, at each instant, the velocity vector of any point. Oftheframeisorthogonaltothestraightlinejoiningthispoint and the ICR. Controlling the vehicle position in the field impliescontrolling the two-dimensional l
21、ocation of the ICR,which may be achieved by specifying the direction oftravel of two points of the vehicle. To get experimentalresults with the 4WS system, a simple controller thatcontrolstwosteeringpointswasimplemented,oneatthefront end and one at the rear of the vehicle. The 4WS isthen utilized to
22、 minimize the distance to the desired pathforboth steering points independently as indicated inFig. 5. This approach with two independent controllersallows us to switch between 2WS and 4WS withouthaving to change the controller structure. As front andrear controllers are identical so without loss of
23、 generality, the description here is focused on the front steeringcontroller. Its control objective is to minimize theperpendicular distance to the path df. The sign of dfindicates the side of the path on which the steering pointis located. From df it calculates a commanded directionof the front steering point (FSP) relative to the vehicle f , using:
