Makarand Karbhari

Design of Actuators for a 4R Robotic Arm

---

Challenge- To determine the dynamic torques required for the motors in a wheelchair mounted 4R robotic arm.

Solution- The dynamic torque analysis requires a much deeper understanding of the various robotics concepts such as forward and inverse kinematics, velocities and static forces, trajectory generation, etc. This makes our desired results to be lot complicated if performed by hand. So, I used the MATLAB’s Robotic System Toolbox (RST) to perform this analysis.

Result- The dynamic torques at each joint of the manipulator were determined using the MATLAB’s RST for the desired positions based upon the application.

• Skills: Matlab
• Date: December 2019
• Category: Robotics

Close

Fixtures Design and Manufacturing

---

Assisted the design team with the generation of complex 3D part drawings and assembly of the generated parts using SolidWorks.

Generated and reviewed the technical content of drawing, engineering specification and reports and other documentation associated with mechanical design.

Assisted the shop floor manager to take corrective actions to improve safety and quality of the product. Gained knowledge about various manufacturing processes such as CNC, VMC Machining, Casting, Welding and Grinding.

• Skills: SolidWorks
• Date: December 2016
• Category: Mechanical Design

Close

Snatch Block Assembly

---

Performed machine design calculation for designing snatch block assembly for load capacity of 80kN. Prepared final design report, CAD model, traveler and engineering drawings of the assembly and final components.

• Skills: SolidWorks
• Date: February 2019
• Category: CAD Modeling, Mechanical Design

Close

VGG16 Model using Transfer Learning

---

In this assignment, I was using a pretrained network and apply it to new dataset. A pretrained Neural network called VGG16 was used for this purpose. 4 different class of test images were provided and the neural network that I created was able to detect these classes successfully with testing accuracy of 73.75%.

• Skills: Python, TensorFlow
• Date: November 2019
• Category: Neural Networks, Deep Learning

Close

Bicycle Drivetrain Assembly

---

Designed a model and assembled a bicycle sprocket assembly with appropriate geometric Dimensioning and Tolerances.

• Skills: SolidWorks
• Date: April 2020
• Category: Part Modeling and Assembly

Close

Object Recognition using Neural Network and Colors

---

Challenge- Different objects requires different grasping points based on their geometrical properties and this can be done if we identify them correctly.

Solution- Object identification can be done with various approaches like with the help of barcodes, RFID or vision systems. Based on the constraints and dependencies in our application I identified the object with the help of a camera using the combination of YOLO V3 algorithm for object detection and a color-detection algorithm to classify the object in order to get the classified object’s properties from the metadata.

Result- This algorithm was tested on different colored objects which were attached with different colored papers as well as multiple colors were attached on the single object and based on the sequence of colors the object was identified correctly in a structured environment.

• Skills: Python, Matlab, Webots
• Date: June 2020
• Category: Computer Vision, Robotics

Close

Position Determination using Monocular Camera

---

Challenge- To get the position and orientation of the object in a 2D using a monocular camera.

Solution- In this experiment I achieved this by using a webcam and LabVIEW software. The object can be located with the help of three co-ordinates i.e. by measuring distance in X, Y and Z direction. In our case I assumed the camera to be at the fixed distance from the object plane which means my distance in Z direction of rotation i.e. along Z direction.

Result- 4 different objects with their position and orientation are determined using this algorithm.

• Skills: LabVIEW
• Date: September 2019
• Category: Computer Vision

Close

Evaluating Camera Transformation

---

Challenge- To get the transformation of the camera with respect to end-effector of the robotic manipulator.

Solution- Various transformation techniques were investigated. Optimal technique was selected based on the design constraints for the camera.

Result- The intrinsic and extrinsic parameters for the camera were calculated. The iteration technique was used where a checkerboard was mounted on robotic arm’s end-effector and moved through to obtain different posses. The difference between the position of detected checkerboard and checkerboard projected into the camera image is minimized to find the transformation. Min ||𝑃𝑐𝑎𝑚(𝑑𝑒𝑡𝑒𝑐𝑡𝑒𝑑) − 𝑃𝑐𝑎𝑚(𝑝𝑟𝑜𝑗𝑒𝑐𝑡𝑒𝑑)|| ≤ 𝑆𝑒𝑡 𝑣𝑎𝑙𝑢𝑒.

• Skills: Matlab
• Date: October 2019
• Category: Camera Transformation

Close

Auto-Simulator

---

Challenge- Integrate the external hardware using the CATIA Dynamic Behavior Modeling,with the help of Arduino controller.

Description- Steering wheel joystick is used to apply brakes/steering inputs. Depending upon the control algorithm control actions are performed.

Result- The expected output of this model deals with the behavior of actuators and motors. It targets the crucial area of Systems Engineering like Testing and Virtual validations.

• Skills: 3DEXPERIENCE DELMIA
• Date: August 2017
• Category: System Engineering, Dynamic Behaviour and Modeling

Close

Reverse Engineering

---

Challenge- To generate a mesh from a 3D point cloud data and to create a Class A surface by using curve-based approach.

• Skills: 3DEXPERIENCE CATIA
• Date: February 2018
• Category: Curve Based Design

Close

Automated Areca-nut Tree Climber and Cutter

---

Challenge- To design and manufacture a prototype of cost-effective and easy interactive machine which can climb the areca nut tree and to cut the fruits attached to the tree.

Solution- A mechanism which considered the varying diameter of the trunk of the areca nut tree was developed. Also, keeping in mind the robustness of the machine static analysis results were obtained with the help of simulation software.

Result- The machine was able to successfully climb the tree with 33% reduced time that would take an average skilled person to climb tree. Also, the machine cost was less by 20% compared to wages of the person who climbed the trees per acre.

• Skills: 3DEXPERIENCE, SIMULIA, Arduino
• Date: May 2018
• Category: Prototype Design and Manufacturing, FEA Analysis, DFA, DFM

Close