1. Less computationally intensive fuzzy logic (type-1)-based controller for humanoid push recovery IIIT-Allahabad Reference paper: Vijay Bhaskar Semwal, Pavan Chakraborty, G.C. Nandi, Less computationally intensive fuzzy logic (type-1) based controller for humanoid push recovery, Robotics and Autonomous Systems, Available online 16 September 2014.

2. Point to covers: Why Fuzzy? What is Humanoid Push Recovery? Why Bipedal? Scientific Investigation of Push recovery Fuzzy set, member ship function, rules Fuzzy Logic Controller Performance Result Conclusion References

3. Objective- Learning based model Developing a mathematical model of a bipedal robot for push recovery is extremely difficult task due to: inherently unstable architecture, higher degree of nonlinearity and freedom hybrid dynamics The objective of this study is to develop an intelligent controller and to implement biologically inspired push recovery for humanoid robots. The objective is to reduce the fuzzy rules and make the fuzzy inference set less computationally intensive and fast. Exploiting the advantages of easy trainability and high generalizability introduced an intuitive fuzzy logic based learning

4. Why Is Fuzzy Logic? Fuzzy refers for uncertainties and imprecision. Fuzzy logic actually captures the fuzziness and vagueness existing in the environment . Many value logic Fuzzy used to developed the more real and low cost solution. The fuzzy inference system takes two crisp values as inputs, fuzzified it, applied number of rules ,and defuzzified the output to convert it into a crisp value.

5. Fuzzy Inference system architecture Fuzzifier Rules Defuzzifier Inference Crisp Input xX Fuzzy Input Sets Fuzzy output Sets Crisp Output

6. Push Recovery Three strategy ( Ankle , Knee and Hip) while F1, F2 and F3 are magnitudes of force Push recovery [1] is the capability of any subject to recover from applied external perturbation with support of other limbs.

7. Why Bipedal? To enter in human like environment 4D (Dirty, Dull, Difficult and Dangerous) . Work human similar environment without and changes in structure i.e. unstructured terrain, climbing of stair case, hazardous environment and narrow terrain These type of system widely used in various real time application like rescues operation, bomb disposal, rehabilitation, mining, hospitality industry etc . The human walk and push recovery is the learning mechanism and it grows with age.

8. Motivation As on date no humanoid robots are commercially available which can negotiate push in real time. However, if humanoid robots are to work in a cluttered environment push is a very commonly experienced phenomenon which we as human can recover from where as humanoids cant. In such cases, the robot could potentially damage itself and its surroundings. Our motivation is a humanoid robot working in a social environment should have some bounded push recovery capability like us. It will make humanoids smart and robust since in real life during working in a unstructured environment some unexpected push may be experienced by the robots.

9. Closed Loop controller

10. Fuzzy Logic based Closed Loop controller

11. Proposed Hierarchical Fuzzy Controller design for humanoid Push Recovery FIS2: Fuzzy Set3:Reaction Small {Roll, Pitch} Average {Roll, Pitch} Large {Roll, Pitch} FIS1: Fuzzy Set3:Reaction Small {Roll, Pitch} Average {Roll, Pitch} Large {Roll, Pitch} Fuzzy Set1-Force {Small (0-5N), Average (4-8N), Large (7-12N} Fuzzy Set2-DoM (Direction of Motion) {Left, Right, Forward, Backward} Strategy Applied {Ankle, Hip, Knee} State (fall, non fall)

12. FIS1 and 2

13. Design The two inputs variables are Force and Direction of Moment (DOM). The corresponding membership function for above two set are following: Fuzzy Set1-The fuzzy value range for linguistic variable Force: Force=Small (x) ={0-5N}, Force=Medium (x) ={4-9N} Force=Large (x) ={8-12N}. Fuzzy Set2-The fuzzy value range for linguistic variable DOM: DoM=Left (x), DoM=Right (x), DoM=Forward (x), DoM=Backward (x)

14. Fuzzy Inference System 2(FIS2) Design. The FIS 2 uses the output of FIS1 as input linguistic variables. FIS2 has output is combination of force and direction applied. Small {Roll, Pitch}, Average {Roll, Pitch}, Large {Roll, Pitch} Fuzzy Set3: defines a linguistic variable Reaction has values Small {Roll, Pitch} Average {Roll, Pitch} Large {Roll, Pitch}. FIS2 have output value in term of whether the robot will able to recover or not and which strategy the robot will apply for recovery. The set for FIS2 output Strategy Applied {Ankle, Hip, Knee} And State {fall, non fall}.

15. Fuzzy Rule 1 Rule 1: action=small roll = max[Force=Small(x),min[DOM=Left(x),DOM=Right(x)]] Rule 2: action=small pitch = max[Force=Small(x),min[DOM=Forward(x),DOM=Backwa rd(x)]] Rule 3: action=large roll = max[Force=Large(x),min[DOM=Left(x),DOM=Right(x)]] Rule 4: action=large pitch = max[Force= Large (x), min[DOM=Forward(x),DOM=Backward(x)]]

16. Fuzzy Rule2 Rule 5: reaction=ankle strategy, not falling = max [Action=small roll(x),Action=small pitch(x)]-(17) Rule 6:reaction=knee strategy, not falling = max[min[Action=average roll(x),Action=small pitch(x)],min[Action=average roll(x), Action=average pitch (x)]]-(18) Rule 7: reaction=hip strategy, not falling = max [min[Action=small roll(x),Action=large pitch(x)],min[Action=large roll(x), Action=small pitch (x)],min[Action=average roll(x), Action=average pitch (x)]]-(19) Rule 8: reaction= falling F/B= max[Action=large roll(x), Action=largesmall pitch (x)]-(20) Rule 9: reaction= falling L/R= max[Action=small roll(x), Action=small pitch (x)]- (21)

17. Experiment al Setup for Humanoid Push recovery Frontal /Backward Plane Subject in indigenous developed HMCD wearable suit

18. Scientific Investigation Of Humanoid Push Recovery

19. Ideal gait curve of different Joints 0 20 40 60 80 100 120 140 160 180 200 -10 0 10 20 30 40 Gait Cycle HIP(degree) 0 20 40 60 80 100 120 140 160 180 200 220 -60 -40 -20 0 Gait Cycle Knee(Degree) 0 20 40 60 80 100 120 140 160 180 200 220 -15 -10 -5 0 5 10 15 Gait Cycle Ankle(degree) (a) hip, (b) knee (c) ankle

20. Results Observed Leg Joint Curve for Right and Left Leg of Right Hand Subject

21. Average Push Force

22. Large Push Force

23. Surface View FIS 1 and FIS2

24. DoM Force Left/Right Forward/Backward Small Small Roll Small Pitch Average Average Roll Average Pitch Large Large Roll Large Pitch Pitch Roll Small Pitch Average Pitch Large Pitch Small Roll Ankle Strategy Knee Strategy Hip Strategy Average Roll Knee Strategy Hip Strategy Falls in frontal plane Large Roll Hip Strategy Falls sideways Falls Fuzzy rule set FIS- 1 and 2 for learning

25. Conclusion Introduces an intuitive fuzzy logic controller for bipedal push recovery. The hierarchical fuzzy logic based controller has been designed to reduce the computational cost incurred by large number of variables. We have designed the hierarchical fuzzy logic controller. It has been tested on the actual data and generalized the hierarchical fuzzy controller for easy trainability. It has been verified that the hierarchical fuzzy system can simplify the complex behavior. Our developed fuzzy inference system is less computationally intensive and able to recover the forces from all the direction. The impact of different magnitude forces on the different

26. References 1. Semwal, Vijay Bhaskar; Bhushan, Aparajita; Nandi, G.C., "Study of humanoid Push recovery based on experiments," Control, Automation, Robotics and Embedded Systems (CARE), 2013 International Conference on , pp.1,6, 16-18 Dec. 2013. 2. Vijay Bhaskar Semwal, Pavan Chakraborty, G.C. Nandi, Less computationally intensive fuzzy logic (type-1) based controller for humanoid push recovery, Robotics and Autonomous Systems, Available online 16 September 2014. 3. Gordon, Sean W., and Napoleon H. Reyes. "A Method for computing the Balancing Positions of a Humanoid Robot." NZCSRSC 2008, April 2008.