ROBOTICS at Carleton – the core in-situ

resource utilisation (ISRU) technology

Prof Alex Ellery

Space Exploration Engineering Group (SEEG)

Dept Mechanical & Aerospace Engineering

Carleton University

RESOLVE PAYLOAD

In-situ resource utilisation (ISRU) – exploitation of extraterrestrial resources - is the key to

development of the space environment

RESOLVE is a US-led mission to lunar south pole Aitken basin

RESOLVE payload (US) to demonstrate in-situ resource utilisation for the first time

Multiple scientific instrument package of 72 kg

Water ice extraction

Oxygen generation through hydrogen reduction of

ilmenite

Canadian rover and drill – industrial team

Canadian science definition team

(i) Soil property extraction during traverse and drilling

(ii) Visual feature extraction to recognise signs of ice

and/or ilmenite

Full teleoperation is infeasible

Autonomous Robotics is core technology for ISRU

A NEGLECTED RESOURCE IN CANADA

Universities are neglected resources – this is particularly the case in space engineering

Kapvik was a 30 kg fully-functional micro-rover prototype for CSA administered by MPB

with Carleton University as technical lead – it was built with a clear PATH TO FLIGHT

Carleton University’s Space Exploration Engineering Group performed the initial design and

specifications of the entire rover and its performance parameters, and performed detail

design (from scratch), manufacture, assembly and test of Kapvik chassis, body structure,

camera/uv pan-tilt system, avionics control architecture, avionics packaging and harness

routing, and major aspects and interfaces of vision and autonomous navigation system for

only $320k over 2.5 years – at all times, we were on budget and on time

It is still working 2 years (the only fully-functional micro-rover platform at CSA) after delivery

undergoing deeper and longer tests beyond its original design specifications

In Europe and US, university teams are valued and trusted to develop and build spaceflight

hardware – scientific instruments (eg. QMC), full spacecraft subsystems (eg. APL), and

innovative technology demonstrators (cubesats – everyone and his dog)

Micro-rovers, manipulators and other micro-systems (eg. micro-penetrators) are within the

capabilities of universities - universities are in principle no less capable with no greater risk

of delivering quality products than companies

We are cheap in financial cost but rich in capability – if you want extreme value for money,

come to us!

KAPVIK MICRO-ROVER

Fundamentally new concept - exchangeable modular chassis - rocker-bogie (built)

and ELMS (detailed design only)

Exchangeability within three minutes by a single human using a single tool

SOIL PARAMETER ESTIMATION

Soil parameters are required to perform automated traction control

To extract soil parameters online, we implemented Instrumented rocker-bogie

chassis for Kapvik microrover

)sin)(sin1())(/[(1)tan(

aasKrec

PETRIE ISLAND FIELD TESTS

WATER ICE DETECTION BY ROVER

MLP neural net models based on Bekker-Wong traction theory

Online extraction of cohesion 3.7 kPa and friction angle of 28 degrees (sandy-

loam)

Online detection of surface water ice and evaporated fluffy regolith (RESOLVE)

Input to traction controller to minimise slip (and power)

ELBOW-MOUNTED PAN-TILT UNIT

STEREOVISION SYSTEM

Vision processing filters extract objects

Disparity map extracts 3D

point cloud for input to SLAM

AUTOMATED ROCK DETECTION ON MARS

LIDAR PROCESSING

To efficiently parse large point clouds from laser

rangefinder as well as classify different terrain for path

planning, neural network classification algorithm was used .

SELF-LOCALISATION & MAPPING (SLAM)

SLAM is necessary to self-locate in the environment – it is first stage of

AutoNav

Object surfaces and terrain estimated using meshing algorithms

Delauney triangulation to self-localise

SLAM using unscented (cubature) Kalman filter (UKF)

SLAM-based autonomous navigation simulation

Path-planning is second phase of AutoNav

Path-planning development with NEPTEC using

ClearPath Husky rover

SLAM USING D* PLANNER

POTENTIAL FIELD-BASED PATH PLANNER

USING WAYPOINTS

POTENTIAL FIELD NAVIGATION ON MARS

Rock distributions for VL1, VL2, MPF & MER-A

MARS ANALOGUE OPERATIONS

Understanding rover operational constraints

Jeffreys Asbestos mine Quebec with large

science team

Pioneer robot used for 3 day field deployment

Mars methane plume source tracking algorithm

AUTONOMOUS VISUAL SCIENCE

There are several approaches to visual feature discovery

In order to explore the information content in texture, we are ignoring colour for the moment – in addition, regolith covering makes colour differentiation unreliable

Shale Schist Igneous

Resolution is critical for extracting useful features – we

have thus far worked at 1m stand-off distances

The use of Haralick parameters offers one potential approach that is computationally

efficient – we are currently investigating this technique applied to rocks (applicable to soil

texture)

Basic image statistics

Entropy measures (one of several Haralick

parameters

Investigation of image pre-processing

(GCOM) matrix

HARALICK-BASED TEXTURE ANALYSIS

GABOR FILTER-BASED TEXTURE ANAYSIS

Canny edge detector Gabor filter

Gabor filter offers greater textural richness – they are direction-sensitive

Human visual processing implements GF

Gabor filter bank of 80 filters for pattern

recognition of textures

High computational complexity

(a) add colour/NIR channels

(b) apply to geological terrain features

BAYESIAN NET CLASSIFIER

Rock classification implements “robotic geology”

AUTOMATED CAMERA CONTROL

Active vision introduces serendipitous science search mode during rover

traverse

Rapid-slew pan-tilt unit implements foveated vision to minimise image

processing

Salience introduces randomised but targetted search

Kalman filter-trained neural net forward model compensates for lack of

camera IMU

This emulates function of human cerebellum

7 DOF Barratt arm control system will validate

Applicable to on-orbit servicing – start of

space-based manufacturing

ON-ORBIT SERVICING

NOVEL DRILLING TECHNOLOGIES

Wood wasp ovipositor

Eliminates need for automated drill

string assembly required in deep

rotary drilling, eg. ExoMars

We have been examining bio-inspired techniques to drilling

MICRO-PENETRATORS

Micro-penetrators may be deployed to the Moon and/or asteroids to

investigate surface and subsurface properties – initial survey missions

LUNAR PENETRATOR TRAJECTORY

ASTEROID IMPACT & DEFLECTION

ASSESSMENT

We are proposing a 3 kg micropenetrator with a minimal suite of micro-

instruments for ESA’s AIM spacecraft to measure debris impacts

generated by US DART

Seismometer is central for internal geophysics

to determine internal consistency

Temperature measurements

for thermal inertia measurements

Accelerometer is GP instrument

This provides basic measurements of asteroid surface and

Subsurface as precursor to ISRU and/or impact mitigation

VIFFING MICRO-PENETRATOR FOR

ENCELADUS

CONCLUSIONS

Canada has an opportunity to participate in the first ISRU mission - RESOLVE

Once demonstrated, ISRU will become more attractive and accepted

ISRU represents the “next big thing” in space commerce

Key concept – robotic factory represents a “mass amplifier” that dwarfs the limits of launch:

it amplifies by producing ~10^5 times its own mass in product

We are interested in the application of 3D printing to ISRU

Key is to match material manufacture to

printing technology

One possible space market is the

manufacture of simple space modules to

construct a space sunshield (independence from

human space programme)

Our business to make robotic ISRU possible

We are always in search of R & D funds!