CS/EE/ME 75(a)

Oct. 16, 2019

Today:
• Teaming & Projects!
• Systems Engineering Concept: Structured Design Artifacts
• GoTChA Chart
• Objective Tree
• System Architectures
• Homework
 Drive-o-Copter
Principle: primarily a driving machine which can “hop” or fly as needed
• Solves dust problem by ground transit
• “Easy-Swap” chasses
• ConOps: 8 km travel, 12 hops, 1 hour autonomous operation

4WD “Swerve” Steer version

Hopping Performance:
1. 8.7kJ per hop
2. 2.84% Battery drain per hop
Power
Draw

3. 10.3m/s forward speed on flight

Hopping to a higher-level Platform

 Drive-o-Copter Project Goals
(for students at 6+ units)

1. Design/Build specialized Urban Circuit Version of the

Drive-o-copter
2. Finalize Sensor Suite
3. Bring up JPL autonomy package on Drive-o-Copter
4. Take-off and landing control
5. Test Extensively
6. Stretch goals
• Hybrid Locomotion Planning

Resources: Arnon Lowenstein (lewinstein@gmail.com), Drew Singletary (asinglet@caltech.edu), Anushri Dixit

(adixit@caltech.edu) , Amanda Bouman (abouman@caltech.edu)
 Drive-o-Copter Project Goals
(for students at 6+ units)

1. Design/Build specialized Urban Circuit Version of the

Drive-o-copter
• Finalize design of new chassis
• Complete computer & electronics design
integration (where located)
• Determine battery location & wiring
2. Build (largely 3-D print) prototype with dummy loads
for sensors
3. Test operation

Resources: Arnon Lowenstein (lewinstein@gmail.com), Drew Singletary (asinglet@caltech.edu), Anushri Dixit

(adixit@caltech.edu) , Amanda Bouman (abouman@caltech.edu)
 Extreme Localization
Assume: A Total Station is used to localize
the robots on the first “leg” of their operation
in tunnel/bunker/cave
1. Investigate: How can we “extend” the accuracy of
the Total station deeper into the challenge
environment
• Multi-leg total station?
• Radio beacons that connect to total station?
• How to automate
• How to deliver
2. If feasible, Design/build prototype

Resources: Joel
 Automated RC Car Project
Goals
1. Redesign Super-structure so that it is less top heavy
2. Add Sensors to wheels for “odometry”
3. Build 2nd car copy
4. Develop odometry estimate using the wheel sensors
5. Improve steering control at high speeds and during backup
6. Stretch goals
• Add sensor to suspension
• Develop whole body estimator
• Develop high speed navigation based on “perception aware” planning
principles
Resources: Jake Ketchum (jketchum@caltech.edu) , Anushri Dixit (adixit@Caltech.edu), Nikhilesh Alatur
(nikhilesh.alatur@jpl.nasa.gov)
 Automated RC Car Project
Goals
1. Redesign Super-structure so that it is less top heavy
• Study the current RC car to understand the key components
• Redesign the placement distribution of sensors, computers, communication,
and possibly battery.
• 3D print and build new prototype structure
• Assemble new “mock-up” of redesigned car. Build a new copy, or reconfigure
the old car?

Resources: Jake Ketchum (jketchum@caltech.edu) , Anushri Dixit (adixit@Caltech.edu), Nikhilesh Alatur

(nikhilesh.alatur@jpl.nasa.gov)
 Today ~Feb. 10 ~Aug. 10
Timeline
CS/EE/ME 75(a,b,c)
~ Dec. 5 ~ Jan. 8 ~ Feb. 10

Q1: Q2: Q3: Summer:

CS/EE/ME 75 SURFs?

Plan Prototype/Test Integrate

Capability Freeze Urban Cave

Demo Circuit Circuit
 Structured Design Method(s)

Recognition of a Need
Design is a
process that
has a general
Problem Definition
structure.

Solution Generation

Analysis & Optimization

Prototyping
 Structured Design Method(s)

The System Engineering Process should have structure

• Structured flow of the problem solving process
• Structured components of each step
• Structured design methods for each step
• Structured Artifacts to capture, archive, and disseminate the
steps of the design process
 Structured Design Method(s)

Recognition of a Need

• Clarification of Objectives
Problem Definition • Establish Function
• Set Requirements
• Generate Alternatives
Solution Generation • Evaluate Alternatives
• Create Details

Analysis & Optimization

Prototyping
 Structured Design Method(s)

Recognition of a Need

• Clarification of Objectives
Problem Definition • Objective Tree
• GOTChA chart
• Establish Function
Solution Generation • Function Diagrams
• System Architecture
• Set Requirements
• Specifications
Analysis & Optimization

Prototyping
 (Structured) Design Artifacts
Main Idea:
• Capture key steps of the design process
• Communicate intent.
• Provide directions to designers, developers, testers
• Different artifacts needed for different users and process stages

Users
• Stakeholders: the people affected by the system design &
deployment
• Management
• Design, Development, and Test Teams
Objective Tree

Why

How
 GOTChA Charts
(for developers and managers)
GOTChA :
• Goals • Use a Quad Chart structure
• Objectives
Goals: high level description Technical Challenges: list
• Technical of main goal(s) and desired expected difficulties
Challenges results of the project/effort • Challenge 1
• Challenge 2
• Approach • Challenge 3

Objectives: list of objectives Approach: list of proposed

required to meet goal methods to realize
• Obj. 1: <obj > < date > goals/objectives
• Obj. 2: <obj> < date > • Act. 1: < build a model >
• Act. 2: < simulate the model >
• Act. 3: < design experiment
Learning to Fly
Anima Anandkumar (animakumar@gmail.com), Joel Burdick (jburdick@caltech.edu, 626-395-4139), Soon-Jo Chung (sjchung@caltech.edu), Yisong Yue (yyue@caltech.edu),

Need/Justification:
• Vehicle performance can be strongly affected by nonstationary and variable effects in its
surrounding fluid flow
• Current practice relies upon exhaustive and expensive wind-tunnel testing to refine control
system design and performance.
• Rigorous nonlinear, optimal, stochastic control techniques that incorporate learning could
improve performance, and lower wind-tunnel and flight testing costs.

Objectives: Based on theoretical, simulation, and experimental investigations, this effort will
demonstrate in the CAST center at Caltech the ability for autonomous fixed wing aircraft and
rotorcraft to safely and stably incorporate real-time learning of changes in environmental and
system dynamics. We will build and demonstrate this capability through 5 demonstrations,
and quantify the efficiency, robustness, and effectiveness of the techniques.

Approach: The proposed program will incorporate the following activities. MILESTONES & SCHEDULE: Proposed 2018/2019 Budget: $535k
1. Theory & Algorithms: We will explore and compare multiple approaches to dynamic model TASKS/MILESTONES 2018 2019 2020
learning based on Gaussian Processes, Koopman Spectral Techniques, and Reinforcement
learning. For safety and stability we will investigate and compare new techniques in Control 1. Develop/improve Test-Bed
Barrier Functions, safe optimization, and nonlinear stochastic optimal control and stabilization. 2. Gather flight data
New techniques that blend the novel learning and nonlinear control methods will be developed
3. Learning Theory
and compared.
2. Simulations: We will develop strip-based aerodynamic models of the CAST flying ambulance to 4. Safety/Stability/
vehicle to serve as a starting model for the learning processes. 5. Control theory
3. Experiments: We will instrument and gather data from a fixed wing and multi-rotor craft
6. Real-Time Optimization
operating in unsteady flow conditions in the CAST center at Caltech.
4. Demonstrations: Demonstrations are planned every 6 months as an integrating focus and as a 7. Demonstration preparation
procedure to evaluate progress, and evaluation methodology. M1 M2a M2b M3a M3b
5. Knowledge transfer: technical papers, copies of the prototype code and algorithms used in Q Q Q Q Q Q Q Q H H
CAST demonstrations, data resulting from simulations and tests, seminars at Raytheon Missile, Research Actuals/Budget 1 2 3 4 1 2 3 4 1 2
exchange of student interns and researchers with Raytheon Missile. or Budget Proposal
106
k 85k 86k 86k 86k 86k 176 176
k k
Function Diagram
 System Architecture
What is an Architecture:
• A model/structure of the system
• Properties of the various elements involved in the system
• Relationships between the various elements
• Behaviors and Dynamics of the various elements
• Multiple Views of the system (from energy usage, information
usage
 System Architecture
Requirements for an Architecture:
• The objects/elements of the system can be modeled (possibly as their
own systems)

• System can be broken down into small systems (hierarchy)

• Can be considered at various levels of abstraction
 System Architecture
• Interactions with environment and other systems

• Interfaces between components

• Socio-Technical Aspects
• Behaviors and Dynamics of the various elements
• Multiple Views of the system (from energy usage, information usage
 System Architecture
• Multiple Views of the system (from energy usage, information usage
 Homework
Individual Tasks:
• Get an account on GitLab: gitlab.robotics.caltech.edu. Joel will then assign permissions
• Get a slack account
• Ask to join caltechcseeme75.slack.com

Team Tasks: (all unit levels)

• Develop a GOTChA chart for your project.
• Develop an objective tree for your project
• Create a Team project page on the course wiki
• Propose to Joel a separate 1-hour/week team meeting time

Team Tasks: (6+ unit level)

• RC Car: meet with Jake to get CAD models, etc. Make new GitLab project directory
• Drive-O-Copter: Meet with Arnon, Luis Pabon, Malcolm Tisdale to get CAD models/update
• Extreme Localization: Organize research sub-teams for different localization technologies