Caterpillar Tech hackathon - PS

The document outlines three problem statements for engineering projects related to depth estimation, visual inspection, and DEF tank quality control. The first statement focuses on developing a real-time monocular depth estimation system for edge AI devices, the second on creating an AI vision system for defect detection in components, and the third on designing a DEF tank that measures quality and controls filling. Each problem includes eligible departments, expected solutions, and constraints for development.

Uploaded by

innovations

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

17 views

Caterpillar Tech hackathon - PS

Uploaded by

innovations

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 5

Problem Statement 1:

Problem Statement Title: Real Time Monocular Depth Estimation on Edge AI

Eligible departments: Electronics & Communication, Electrical & Electronics,

Instrumentation Engineering, Robotics, Artificial Intelligence – Machine
Learning, Data Science, Data Analytics, Information Technology, Computer
Science and Mechatronics only.

Problem Statement Description:

Monocular Depth Estimation is the task of predicting the depth value (distance
relative to the camera) of each pixel given a single (monocular) RGB image.
This depth information can be used to estimate the distance between the
camera and the objects in the scene. Often, depth information is necessary for
accurate 3D perception, Autonomous Driving, and Collision Mitigation Systems
of Caterpillar vehicles. However, depth sensors are expensive and not always
available on all vehicles. In some real-world scenarios, you may be constrained
to a single camera. Open datasets like KITTI/NYUv2 can be used. Solutions are
typically evaluated using Absolute Relative Distance Error metric. Based on the
distance between the camera and the object (Cars/personnel), operator
needed to be alerted visually using LED/Display/Audio warnings.

Expected solution & Tools that can be used:

Use either neural networks or classical algorithms on monocular camera
images to estimate the depth. The depth estimation should be deployable on
cheap edge AI devices like raspberrypi AI KIT
(https://www.raspberrypi.com/products/ai-kit/) but not necessarily on
raspberrypi.

Data Set/Reference Date KITTI Dataset can be

used https://www.cvlibs.net/datasets/kitti/
Prototype requirements (If Applicable) Solution should be deployed in edge AI
devices like
https://www.raspberrypi.com/products/ai-
kit/ https://www.beagleboard.org/boards/beagley-
ai https://www.ti.com/tool/SK-TDA4VM
https://robu.in/product/nvidia-jetson-orin-nano-super-developer-kit/
but not limited to these Edge AI devices.

Problem Statement 2:

Problem Statement Title - Create Vision system to identify and display

defects on components

Eligible departments: AI Data Science, Mechanical, Production, Manufacturing,

Aeronautical, Industrial Engineering, Automobile, Metallurgy, Mining and
Mechatronics Only
Problem Statement Description:

Visual inspection on Paint aesthetics, Weld aesthetics, gap measurements are

currently manual posing issues on repeatability and dependence on manual
processes. Technology solution needed to resolve the industry pain.

Expected solution:

Create an AI based Vision system with built in Deep Learning model to detect,
identify and display defects on components which can qualify Paint
appearance, Weld appearance, assembly quality and machining quality.

Problem Statement 3:

Problem Statement Title: DEF tank with mechanism to measure DEF quality &
control DEF filling

Eligible departments: Mechanical, Production, Manufacturing, Aeronautical,

Industrial Engineering, Automobile, Metallurgy, Mining and Mechatronics Only
Problem Statement Description:

An ideal DEF /AUS 32 solution used for NOX reduction should meet ISO 22241-
1 standard and is typically 32.5% Urea in de-ionized water. However, many DEF
solutions available in the mass market are either adulterated or not meeting
this specification. This would impact tailpipe emissions, engine/machine
operations & reduce the capability of after treatment systems. DEF systems
are already equipped with the DEF quality sensor however the sensor would be
able to sense only after the filling is completed when Key On. By this time, the
tank would already be filled with adulterated urea or would adulterate the good
DEF that’s already inside

Problem description and Expected solution:

Design and develop a DEF tank that would measure the quality of urea and
equipped with a mechanism to control the filling of the DEF before it goes into
the tank.

The solution should indicate the machine operator regarding the

condition of DEF quality before the tank is filled. The mechanism should
be quick enough that the filling rate/time should have little of no impact.
Design should be structurally durable to withstand harsh environmental
conditions the DEF systems would be exposed to.

Constraints:

Proposed design should be user friendly & capable of mass market

production
Understand & ensure the proposed solution should meet regulation
requirements pertaining to Urea filling
Standard filling gun specifications to be considered.

Expected Outcome:

The proposed design solution to meet design/functional requirements, be

cost effective, manufacturing & customer friendly
Simulations to prove structural durability
Working prototype that could demonstrate the expected capabilities
CATERPILLAR CAREERS

By becoming a Caterpillar talent community member you agree to our

Terms of Service and consent to Caterpillar and our affiliates
collecting your personal data in accordance with our Privacy Policy. In
addition, you consent to our use of website tracking methods,
including cookies, and the safe and secure transmission of your
personal information. For more information on how Caterpillar treats
your personal information and protects your privacy, please refer to
our Privacy Policy.