Hội nghị toàn quốc về Điều khiển và Tự động hoá
Real distance and measured distances.
distance is maintained.
shows the real distanceand various measured distance.
shows generalswim function .
i i i i i
A t K Am ft t
is the angle of i
tail motor, K
is amplitude factor,A
is amplitude, f is frequency of caudal fin,
isphase delay of i
is deflection angle forslow and quick turn. We use 10 degree maximumamplitudes of angle and 35 degree phase delays forgeneral swim. Swim frequency is 0.5 Hz.
When the approaching angle of the robotic fish is θ
and the swing angle of head is Ai, the fish robot readsreal distance as dreal and long distance as dlong.Therefore, it is necessary to estimate the exact realdistance from the robot to the obstacle. A fuzzydistance estimation method is used to find the exactdistance regardless of the swing of its head. The meanvalue of distance data of one period from the past half to the future half period is defined as a real distance.However, it is impossible to use future distance datain a real time system. Thus, the previous distance of one swing period is used get distance estimation. Thisdata is mean distance of half period ago. Therefore, amoved distance in the past half period should beadded to get a current distance. The sampling time is20 ms, and the swing period is 2 seconds are appliedfor the fish robot.The flowing procedures are proposed to get a fuzzydistance estimation:-
Calculate the mean value of distance data (D
)for past one swing period.-
Calculate the mean value of past ten samples atcurrent (D
) and past (D
) one swing period ago.-
Calculate the distance change for half swimperiod (O
) with O
Calculate a compensation factor O
which iscalculated by a fuzzy logic system.The inputs of fuzzy system are distance data Dm,distance change for half swim period Op and thedirection information D. The ANFIS inference isemployed to train the distance estimation.
Proposed Method For Fuzzy DecisionMaking on Direction Changes
Robotic fish probably encourages serious difficultiesavoiding obstacle collision due to various kinds of obstacles in the water. In order to approach naturaland smooth movement for robotic fish concerns thedetecting and recognizing obstacles as well as theavoiding collision, we propose a fuzzy decisionmaking on trajectory direction changes. In theexperiment, the measured distance from sensor toobstacles and the possible existence of obstacles aretwo essential input data for the fuzzy algorithm.
Establish the upper and lower margins for the inputs
The assignment process to define the upper and lowermargins can base on some algorithms or localoperation as well as intuition. It also derived throughunderstanding and prior knowledge about the systems. For example, the upper and lower margins of theCenter range are 15
, respectively. The upperand lower margins of the Long range are 20
B. Defining fuzzy membership functions
The choice of membership function will help definethe output surface that is based on the combination of the multiple surfaces. Because of insuring smoothtransition among these surfaces, membership functionsmust be even, same type and input domain must beequally divided. The range of obstacle position (O.P.)is divided into three membership functions of Left,Center, and Right which are mean that the intensitiesof obstacles in left, center, and right side are higherthan any other ranges, respectively. The range of obstacle distance (O.D.) is also separated into threemembership functions of Short, Medium, and Longexpress distances from a sensor to the obstacle. Theoutput for the direction change is divided into fivemembership functions of LB: Turn Left Big, LS: TurnLeft Small, Center, RS: Turn Right Small, RB: TurnRight Big.
. demonstrates the two fuzzificationsof inputs for the obstacle distance and obstacleposition and an output for the direction change.
Fuzzy membership functions of obstacle position.(b)
Fuzzy membership functions of obstacle distance.