Human Sensing System for Safety Agricultural Robot

Japanese version is here.

M.Monta, N.Kondo and K.Nakatsuka

Department of Agricultural Engineering, Okayama University

Recently, several agricultural robots, especially for harvesting, have been developed in Japan. It is considered that robots can autonomously achieve the task in the greenhouse and the field in the future. However, another stage when robots and human workers do cooperative works in the same working environment will exist before the stage of complete robotization comes, because robots do not always achieve all farm workings easily. For example, pruning operation is performed by using human experience and sense, forecasting the growth of plants. Therefore, it will be difficult for robots to achieve these operations because their skills are inferior to that of human being, even if the latest technology is utilized for control of robot. In the stage when robots and human workers achieve their tasks in the same environment, human cooperative robot system will be necessary.


Industrial robot in factory is surrounded by iron barriers in order to avoid accidents. The robot will be controlled to stop by emergency system if human worker enters the working area of robot. However, it seems that it is difficult to apply this system to agricultural robots since agricultural robots have to travel in the greenhouse and the field. Furthermore, this system is not always efficient system from a view point of working efficiency. Robot has no choice but to work or stop. Therefore, human cooperative robot system has to be equipped with both safety and efficient elements. In other words, robot who can recognize its circumstance and can decide its motion by itself will be required.


Our research group supposes a safety and efficient robot system which consists three sub-systems. The first one is human sensing system to detect existence of human being and human motion including position, moving direction and velocity. The second one is a system to calculate degree of danger based on information from the sensing system. For example, the degree of danger will be very high if relative distance between robot and worker is short and relative velocity is high. Finally, robot motion is decided by the third system which outputs the information to the robot, in proportion to the degree of danger, such as acceleration, deceleration, stop and avoidance. In this presentation, human sensing system will be described.


The human sensing system consists of four pairs of ultrasonic sensors and infrared sensors which are arranged at an interval of 90 degree, motors which scan these sensor units over all directions in two dimensional space parallel to the ground. Ultrasonic sensors are used to measure distance to human body. Distance can be calculated by measuring propagation time of the sound wave reflected by the object. Each sensor scans 180 degree in 1.3 sec. Several distance data can be detected to one human body because ultrasonic sensor collect a data at an interval of 5 degree and the human body has its width. Therefore, a point which represents the human position is defined by calculating a center point of collected data in two dimensional space. Motion of human body can be traced when representative points are calculated, and moving direction and velocity of human body can be obtained based on co-ordinates of each representative point and rotational angle of motor. Infrared sensors are used to detect the existence of human being and to measure the human directions. This sensor can detect the motion of object which radiates infrared ray and can output analog voltage. Its sensitive wave length range is from 5 to 14 um and its directivity is 15 degree. This sensor is suitable for detection of human body because human body radiates from 9 to 10 um infrared rays. Four pairs of infrared sensors are rotated by motor at a speed of 11 rpm. Reproducible shapes of output voltages can be obtained by rotating sensors when there is no human being, although output voltages change according to the directions of sensors. A transformation of output voltage appears in the direction of human body when a worker exists within the detecting area of sensor. The peak can be extracted by subtracting the shape of output voltage when there is no human body from that when there is a worker. Furthermore, human direction is distinctly detected by differentiating the wave shape, even if the transformation of wave is small. Human detecting experiments were performed by using the sensing system. From the results, it was observed that this sensing system could satisfactorily detect human motion though error in represent of human position was occurred due to the directivity of ultrasonic sensor and the body parts which reflected sound wave.

Expriment results obtained by the human sensing system are showed as follow.


Agricultural Machinery Labo. of Okayama univ. WWW Server "Mama".