Embedded Files
Push Recovery :
We as human being can negotiate with external push up to certain extent reactively. Different people have different push recovery capability. The acquired push recovery capability, therefore, is based on learning. However, the mechanism of learning is not known to us. Researchers around the world are trying to explore this mystery through developing various models and implementing them on various humanoid robots. All the models based on conventional mechanics and controls have inherent limitations. We believe appropriate computational model based on learning will be able to effectively address this issue. I will also discuss about how we can frame this problem as hybrid system. Below figure 1 is describing how human negotiate with different external push reactively. The second fig. is the model with 6-DOF and last one is the architectural view of Humanoid's bounded push recovery during interaction with clutter environment.
Mobile Embedded Accelerometer to capture the human push recovery behavior:
Our Experiment for scientific investigation of human push recovery, The below figures representing the different phase of gait.
We designed the wearable accelerometer embedded mobile phone suit named HLPRDCD (Human Locomotion and Push Recovery Data Capture Device) to capture the data of different joint angles change (hip, knee and ankle) which is the manifestation of locomotion/push recovery.
Accordingly we have collected extensively humanoid push recovery data using our innovative idea of exploiting the accelerometer sensor of smart phone. Through our experiments we have studied the human push recovery by fusing data at feature level using physics toolbar accelerometer of android interface kit. The subjects for the experiments were selected both as right handed and left handed.
It is an improvement of our earlier developed HMCD suit( human motion capture device). Lot of researches are going under the leadership of Prof.G.C.Nandi and
research scholar Mr. Vijay Bhaskar Semwal with dedicated m.tech. team working for more accurate study of human push recovery.
1- Subject one wearing the device for Hip movement. 2- Subject two during stance phase with our sophisticated mobile embedded device wear on left knee.
Our Model:Bipedal Locomotion sub phases of gait cycle
Phase Wise decomposition of Human Locomotion
Broadly we sub divided human gait into 8 sub phases
Curves of Left and Right Hip and Knee respectively during different sub phase of gait:
Our consideration from standard record for gait sub phases
Stance phase:
Initial Contact [0 – 2]% IC Not considered
1. Loading Response [2 – 10]% LR
2. Mid Stance [10 – 30]% MST
3. Terminal Stance [30 – 50]% TS
Swing Phase:
4. Pre Swing [50 – 60]% PSW
5. Initial Swing [60 – 73]% ISW
6. Mid Swing [73 – 87]% MSE
7. Terminal Swing [87 – 100]% TSW
The top most curves in red is our equation generated plot and blue curve for HOAP2 model.
We considered 51 sample and joint angle in degree and HOAP2 data’s joint angles are in digital count with respect to time.
Right Hip
Left Knee
Our Hybrid automata equation for different sub phases of Gait;
1- Right Hip equation for different sub phases
2- Left Hip equation for different sub phases
3- Right Knee equation for different sub phases
4- Right Knee equation for different sub phases
For More Details see here
We analyze the human push recovery using several experiments using HMCD as well as HLPRDCD as manifestation of knee, hip and ankle joint angle change. The mobile phone based data collection is very convenient and accurate compared to the potentiometer based HMCD. Using LVQ we observed that the push recovery capability depends on many factors including age, sex, weight ,height, ambidextrous, race etc.
Our Publications:
Vijay Bhaskar Semwal, Pavan Chakraborty and G. C. Nandi. "Less Computationally Intensive Fuzzy Logic (Type-1) Based Controller for Humanoid Push Recovery" Robotics and Autonomous Systems,Elsevier, 2014 (ROAS)(Impact factor 1.05- H5 factor-37). (Published online on September 16, 2014).(PDF).
Vijay Bhaskar Semwal, Manish Raj and G. C. Nandi. "Multilayer Perceptron Based Biometric GAIT Identification " Robotics and Autonomous Systems,Elsevier, 2014 (ROAS)(Impact factor 1.05- H5 factor-37). (Published online on November 21, 2014).
Biologically-Inspired Non-Linear Hybrid System Humanoid Locomotion through Hybrid Automata
The earlier developed two stage hybrid automata is not perfect representation of human walk as it is combination of discrete or continues phase and the whole human GAIT have 8 stages. Our major contribution is eight stage hybrid automata for large push recovery and various dynamic parameter studies for stable walk model. We developed a controller to verify different stage of human locomotion by using OpenSim data for model 3DGaitModel2354 and lower extremity data. We verified the hybrid automata model using the real human GAIT data for normal person. We identify the importance of the human lower extremity for locomotion and push recovery from large perturbation.
Less Computationally Intensive Fuzzy Logic (Type-1) Based Controller for Humanoid Push Recovery
This paper presents a new type-1 fuzzy logic based controller for push recovery of humanoids. The objective of this paper is to develop an intelligent controller and implement biologically inspired push recovery for such robots. We take two crisp values as input, fuzzify it, then use number of rules and finally defuzzify the output to convert into crisp value. We apply fuzzy rule to our model and simulate it in unstructured environment. We reduce the fuzzy rules and make the fuzzy inference set less computationally intensive, fast and also capitalise on advantages of easy train-ability and high generalization. Our fuzzy logic based controller is able to predict which certain push recovery strategy is required and whether the robot will be able to recover or will eventually fall. The architecture is hierarchal in design. First Fuzzy Inference System (FIS1) is based on two input variables, Force and Direction of Motion (DoM), whose result depends on magnitude of force applied on the body and direction in which body moves as an effect of push. FIS1 is able to tell about small, average and large force in term of roll and pitch effects in body. Now, we use these output as input variables for our FIS2 and predict about which certain push recovery strategy will be applied and eventually robot will be able to recover from push or fall. We introduce term auto leaning in human autonomy for push recovery. We extend Gordon model for balancing humanoid using fuzzy logic and consider effects of roll, pitch and yaw. We extend earlier studies done in field of humanoid push recovery capability when force is applied only one direction to the push exerted from different directions. The novelty of paper is type-1 based fuzzy logic controller for push recovery with less number of variables.
μaction=small roll = max[µForce=Small(x),min[µDOM=Left(x),µDOM=Right(x)]] -(5)
μaction=small pitch = max[µForce=Small(x),min[µDOM=Forward(x),µDOM=Backward(x)]]-(6)
μaction=large roll= max[µForce=Large(x),min[µDOM=Left(x),µDOM=Right(x)]]-(7)
μaction=large pitch = max[µForce= Large (x),min[µDOM=Forward(x),µDOM=Backward(x)]]-(8)
μreaction=ankle strategy, not falling = max[µAction=small roll(x), µAction=small pitch (x)]-(9)
μreaction=knee strategy, not falling = max[min[µAction=average roll(x),µAction=small pitch(x)],min[µAction=average roll(x), µAction=average pitch (x)]]-(10)
μreaction=hip strategy, not falling = max[min[µAction=small roll(x),µAction=large pitch(x)],min[µAction=large roll(x), µAction=small pitch (x)],min[µAction=average roll(x), µAction=average pitch (x)]]-(11)
μreaction= falling F/B= max[µAction=large roll(x), µAction=largesmall pitch (x)]-(12)
μreaction= falling L/R= max[µAction=small roll(x), µAction=small pitch (x)]- (13)
Keywords: Artificial Intelligence, Fuzzy Logic, Robot Dynamics and Kinematics, CRISP set, Fuzzification, Humanoid Push recovery, Machine learning, Defuzzification.
Details Of My work:
The key research challenge is the program the humanoids with mobility and robustness that are similar to human. The major challenge arises when robot works in human environment ( 3D –Dirty, Dull, Dangerous) such as mines, factories, Bomb Disposal and under the sea). These environments are extremely unsafe for human so it is require robot should perform their tasks beyond human capabilities. The future generation of humanoid robot has another challenge as they are expected to co-exist and operate in a human environment. The real challenge for researchers is to program these robots to survive in an environment consists of unknown disturbances. As to understand the human push recovery is required the study of human gait pattern. Any study related to human push recovery can help elderly and disabled people to walk with confidence. The pattern can help for developing the prosthesis leg and analysis of different GAIT pattern. There are certain patterns in human locomotion which is also marked as the unique bio metric identification of particular person. The study using different statistical technique of human GAIT can lead a step for development of Humanoid robot. The classification and pattern reorganization required the depth study of different statistically technique. To analysis the different data for different magnitude push and different GAIT cycle which help to selection of feature by using statistical technique help me to classify different push recovery and to identify different gait cycle.The Humanoid locomotion hybrid automata required strong understand of statistical technique. Different pattern data required data smoothing, data normalization and help for biometric identification of push recovery data.
INTRODUCED TERM(Auto leaning)
I introduced new term auto leaning in human autonomy.During my research sample collection I reached on the very interesting observation that left hand person have constant curve for left leg and vice versa. Above statement conclude that the human structure always observed pressure of push based of their working hand position I introduced term auto leaning for this behavior.
Bipedal Push Recovery Based on Learning
RESEARCH INTEREST
Humanoid Robot,Machine Learning,Fuzzy set,Iinstuitionistic fuzzy sets and Logic,ANN(Artificial Neural Network),Pattern Recognition Wireless Sensor Network, Ad-hoc Network, Vehicular Technology
Software Development,Design and Analysis of Algorithm,Data Structure,Soft Computing,Protocol Development, OOP's concept,Multi agent system for utilization of channel spectrum,Genetic algorithm, PCA, LDA for pattern matching and classification,Hybrid automata design for push recovery.
The primary aim of my research is to develop a Humanoid Robot with push back capability which can resist the sudden push as we human beings observed by different means.There are three strategy for small push ankle strategy, little more heap rotation and if thrust is high to take step one or more then one to come in to balance state using the LIPM(Linear inverted pendulum model) which have constant height CoM and using the orbital energy which is more useful to calculate the liner equation although the human behavior is non linear.By using orbital energy we will calculate the capture point set which could be the set of point which lead for humanoid balance.The robot is learnign based on adabtive learning like us and for understanding or perception using non linear nature by using ANN and fuzzy logic.We making the robot in stable state by maintaing the CoP or CoM .
Specially, the objectives of the work are as follows
To develop a humanoid robot with push recovery capability which can work in environmental which is detrimental for human beings.The scope is not limited with these application
To explore it with fuzzy logic or for machine logic using ANN(Artificial neural network) for learn from real world non liner data with adaptive learning
Page updated
Google Sites
Report abuse