Semi Active Modular Robot for Uneven Terrain Navigation

Partially Actuated at the joints for smart maneuver

• In this robot alternate link joints are actuated.The aim is to use wheel traction for climbing.
• Link actuation is used sparingly to avoid tip-over. Link actuators are activated only when the module reaches a certain threshold angle that is indicative of tip-over.
• A model based control scheme is developed to achieve the desired control action at the link joints.
• The 5-module semi-active robot can climb steps of heights up to 2 times its module length.

Snapshots of the semi-active robot traversing on an Uneven terrain.

A Novel Compliant Modular Robot for Traversing on Uneven Terrains

A Compliant type Robot which uses springs at joints for smart maneuver.

• A 3-module passive robot is developed with compliant joints to avoid tip-over.
• An optimization problem is formulated based on the quasi-static analysis of the robot to estimate the stiffness values for the compliant joints.
• Passive robots are generally more durable.
• The robot was able successfully climb a step of height 17 cm.
• It shows very good terrain traversal capabilities.
• The 5-module variant of this robot successfully climbed a step of height 36 cm.
• The use of springs also helps in some slip reduction due to the effective redistribution of normal forces. It also helps improve climbing efficiency.

Wheel Torque Control of Modular Robots for Uneven Terrain Navigation

Optimizing torque of wheels to reduce wheel slip.

• The prime motivation behind wheel torque control Is slip reduction while climbing step-like obstacles.
• It is achieved by fusing optimal torque controller with a velocity controller.
• The desired torque values are obtained from an optimization problem that is formulated with an objective to minimize traction-to-normal force ratios for all the wheels.
• The numerical and experimental results validate the efficacy of this method. There is a considerable reduction in wheel-slip when this controller is used.
• The robot was made to climb steps from a tiled, carpeted, taped and wet floors during the experimental studies.

Content Categorization, Sentiment Analysis and Data Tagging & Analysis algorithms

Multiple Projects with different targents to achieve.

• Automation of Categorisation of data.

  A large amount of data of client (Comments from various social sites) is collected and is categorised based on the pre-defined tags as submitted by the client. Also prioritising the issues, which is to be focussed. For eg. 'cannot open' will be tagged as 'Error Issues'

• Keyword Inflection.

  Grouping/Clustering of large amount of data (Comments or Search Terms) into relevant groups keeping the misspells also into consideration. Using wordnet dictionary along with lemmatization and stemming algorithm to group the similar data for both Unigram and Bigram words.

• Keyword categorisation.

  Crawling the website of the client to extract the categories and subcategories of the product. And then categorising the search term accordingly. ( Also incorporating the dictionaries and other algorithms for improvising the results).

• OCR error correction.

  When doing OCR recognition of document, many errors are introduced in form of misspell or alternate meaning (alteration of character). The task is to correct the word, according to the context. The trigram based approach is used along with dictionary suggestions to suggest the most relevant alternatives for the erroneous word.

• Long-tail Analysis.

  Based on the pre-criterion and algorithms provided by the client, the data is formed, sorted and clustered into multiple groups for them to analysis. For example the data is sorted on product related search for company having 10 products. And upon sorting we find that top 3 products account for 90% search terms. Then company should focus on rest to grow.

• T-Pie

  As a support for the GUI created for Sentiment Analysis, Clustering and Tagging.

Velocity and Position controlled Simulation of Quadrotor

Controller written for the simulation of Quadrotor

• The prime motivation behind the simulation is the visual servoing of the quadrotor to follow an object.
• It is achieved by fusing optimal velocity controller with camera feedback.
• The desired velocity values are obtained from an error calculation between the desired image and the actual image and fusing it with the interaction matrix.
• To verify and to generate the controller to follow the velocity inputs, simulation is done where the way points are already given and the quadrotor has to go through the points.
• In next step a velocity profile is generated (a continuous curve). Quadrotor has to successfully follow the profile. Complete dynamics of the quadrotor is incorporated in simulation and controller was developed.

(a) Quadrotor following the velocity profile to go diagonally to the cube.

 

(b) Quadrotor going through the given waypoints (0,0,0),(1,0.5,0.75),(2,1,1.5),(3,0.75,1.0),(4,0.5,0.75),(5,0,0.5)

 

Detachable Compliant Modular Robot

A compliant modular robot to detach in individual modules to explore and reattach to overcome obstacles.

• 3 modular compliant robot already prototyped and tested, with increased ability to ascent and descent stairs.
• Passive Robots are robust.
• Increased productivity needed while exploring the planar area. Hence the idea of detaching in separate individual identity came up.
• Mechanism is created to detach and re-attach the individual robots keeping the compliance intact.
• Increased maneuverability as roll-joint is also incorporated.
• Tight space maneuvering is also made possible.
• Lesser time required to explore certain area in Urban Rescue Mission scenarios.

Autonomous Car

Path Planning for autonomous maneuver in lanes.

• Working on the path planning for Autonomous Vehicle, for the smooth manoeuvre in highways. Features like over-taking, following, stopping etc. are being implemented on planner.
• Implementation of various types of center curve trajectories. Testing each curve in various scenarios and then finalising the best one.
• Incorporationg the feature of Traffic Signal behavior to path planning.
• Implementation of obstacle avoidance and changing trajectories accordingly.
• Real life testing of autonomous vehicle and tweaking code, for better results.

Circular Maze Solving Robot

A line follower robot solving the circular maze according to predefined set algorithms.

• Problem statement of Techkriti-2013 (IIT Kanpur's Technical Festival). (2nd Year of College)
• Robot is targetted to reach at the center of the maze in minimum time by following the path, with set of rules.
• Robot is to follow the perimeter of circle it is in, untill a marker comes, and it needed to turn in direction of the marker whenever next turn comes.
• Ground is of green colour, markers of black and end points were of white colour. And robot is expected to detect the same.
• My robot comprised of Atmega32 as controller and LDR as sensors connected to ADC pins.

Maze Solving Robot

A line follower robot following maze, and returning from the shortest possible path in least time.

• Problem statement of Avishkar-2014 (MNNIT Allahabad's Technical Festival).
• Was in organising committee and hence made a working prototype as an example.
• Aim was to explore the maze, its nodes to reach to the final point. And solving the maze to find out the shortest possible path. And retracing the same path as fast as robot can.
• Robot had Atmega32 microcontroller and an array of 7 pairs of IR transmitter-receiver as sensor inputs.
• Standard maze solving algorithm (↑) was implemented with Right-Straight-Left strategy.
• To minimize the runtime, high speed motors (300rpm) was used and velocity was controlled using PID controller.

Image Processing Robot

A robot which identifies the colour of ball and put it in respective coloured pit.

• Problem statement of Avishkar-2014 & Avishkar-2013(MNNIT Allahabad's Technical Festival) with slight changes in final aim.
• Aim was to search the area to find a particular colour of ball, approach it , pick it and place the same in the designated area in the corner.
• Robot had Atmega32 microcontroller with serial communication to laptop to control movement. Visual feedback was provided by web-camera attached to robot chasis and directly connected to laptop.
• Video/Image input and processing was done on MATLAB.(↑)
• With each input image, the defined colour ball was located. Approximate distance between the robot and ball was calculated by the area of projected image of ball, and theta by the deviation from the center of the image. Required velocity & omega was calculated accordingly, which is serially transferred to microcontroller.
• Microcontroller decode the velocity-omega parameters to calculate each wheel's velocity (rpm) and move the robot.
• Velocity was controlled using an open loop controller.
• A similar run of events was done with different colours to pick and place the objects.

Retro Snake Game

Snake game was implemented in 8x8 LED matrix.

• Problem statement of Techkriti-2013. (Annual Technical Festival of IIT Kanpur)
• Aim was to implement a snake game on LED matrix.
• We implemented it on 8x8 matrix, and whole setup costs <300INR.
• Snake has ability to move in the matrix, and was of cyclic in nature i.e. if it enters through one wall, it comes out from opposite side.
• The fruit points were randomly generated, and snake length was increased by 1 unit after taking it.
• The module had two modes, classic and challenge. Classic has no predefined walls, while challenge mode has walls. The current image shows challenge mode with diagonal wall.
• After game completion the module shows the final score.
• It has reset, pause and on-off buttons.
• This won the 3rd prize in the event and special mention for being cost effective.

Home Automation System

IoT based home automation system controlled both by laptop and mobile.

• Problem statement of Avishkar-2013. (Annual Technical Festival of NIT Allahabad)
• Aim was to implement home autmation system, where light bulbs, fans etc. can be controlled via electonic devices, and has functionality of auto-shutdown.
• The module has 2 light bulbs, and one fan. The fan was regulated at different speeds.
• Laptop and the module communicated via serial link and bluetooth.
• Communication between mobile and the module was with bluetooth and DTMF. Thus giving it benefit of being independent of short range and internet.
• Module has auto switch off mode, if there was no activity around. (Idle time was set of 10min)
• Fan speed was also regulated automatically according to the temperature of the room. Lesser the temperature slower the fan be.
• This won the 1st prize in the event.

MATLAB Simulations

Simulations done in order to learn.