Drones Navigating without GPS: The Future of Autonomous Drones with Dr. Tim McLain

it's beneficial even if nobody has

Global position information so and then

flying in and out of a warehouse for

example Amazon warehouse It's

anticipated that in a serious conflict

that GPS would be

[Music]

eliminated welcome to the tech transfer

podcast I'm Dave Brown I'm here with

Professor Tim mlan and we are going to

talk about a UAV technology that is good

for uh maning those situations where

there's no GPS signal so Tim what's

what's the technology and what problem

were you addressing so the technology we

we call it broadly relative navigation U

uh We've applied it to systems just

single aircraft systems but in this case

to multiple aircraft working together

and the basic idea is that if the

aircraft can communicate with each other

then there's information contained in

that communication about how far apart

the aircraft are from each other if

they're in a situation where they don't

have GPS they can exploit that inter

vehicle range measurement to help each

aircraft figure out better where they

are in the absence of GPS okay so this

could be any kind of aircraft probably

I'm picturing small drones and when when

they communicate with each other do they

do they know how far apart they are

because of how long the signal takes to

move yeah that's the idea is that

there's a technology it's called ranging

radio and basically that uh whenever a

communication transmission happens

there's some time of flight um

processing that's done done that gives

basically range information okay so a

group of drones or whatever vehicle are

communicating with each other and

they're also doing some mapping right

how does that part of it work so um in a

GPS denied scenario or even degraded GPS

scenario the individual aircraft can use

onboard sensors such as accelerometers

rate

Gyros uh cameras cameras looking at the

terrain or or uh the environment below

and they can use features in the

environment imagery along with these

other sensors to um basically figure out

how they're how they're going they it's

called a visual odometry they can figure

out how they're traveling over what if

there's uh known landmarks that they see

they can figure out where they are

relative to those landmarks and if those

landmarks are in known locations then

they can figure out where they are in

the world from using camera information

that's great and so I I should note that

we have a patent that's about to issue

on this technology and it it it will

cover among other things this mapping

function when you say that it's uh the

they're noticing like known terrain this

is not necessarily like marks on a map

it could just be features that they

recognize right that's right so it could

be just visual features that they're

that are distinct in the imagery or it

could be that they're known landmarks so

they could be um you know a building a

certain building where U referencing an

existing map the aircraft could know

where it was relative to that okay so

let's go through some scenarios just to

see how this would look in the real

world so one thing that immediately

comes to mind is military applications

if you had drones flying through a

canyon or something they could be they

might lose the GPS signal because of the

canyon walls but they're talking to each

other and mapping the train is that a

realistic thing that could happen yeah

that's I think um a concept in Canyons

of course GPS can be degraded or or lost

altogether um so that's a scenario

another scenario that's perhaps more

likely is jamming of GPS or just allout

obliteration of GPS oh that's

interesting so in a military situation

yeah there'll probably be some kind of

jamming yes yeah so the idea I guess

would be that if you had a group of

vehicles that were flying as a a swarm

or a group um it's important that they

know where they are

and if they were to lose GPS but they

had the ability to localize relative to

the terrain or as they were flying and

by localize I mean know their position

relative to just visual features MH um

that they could uh use this visual

odometry idea to figure out U basically

how they were moving with respect to the

terrain and have some sense of where

they are however uh when you're doing

techniques like this visual odometry you

have sources of error like typically you

would say that the heading error of the

aircraft that the the knowledge of the

aircraft uh the knowledge of the

aircraft's uh direction that it's flying

is um is not as precise as you would you

would hope and as it flies doing this

visual odometry it accumulates

significant amounts of air and so it can

drift off course basically so if you

have multiple aircraft that are drifting

off course so to speak

but they're able to communicate with

each other you can basically get rid of

some of that drift or a significant

amount of that drift error by from these

inter vehicle range measurements

furthermore if any one vehicle knows

where it is all of the vehicles can know

where they are by communicating um their

uh their uh relative pose or relative

range or relative bearing all of that

sort of information can be

exploited and and um and they need to

also be able to communicate

um where they

are uh at the time of their

communication so in a scenario like the

canyon one we're talking about for a

second could you have like if it's a

group of drones could you have one of

them flying much higher so it's not even

in the canyon but it it has a clear line

to communicate with them would that

solve the problem yeah in in some

situations uh an aircraft flying high

above the canyon would have GPS and

could provide Global Information to the

aircraft below but this concept it's

beneficial even if nobody has Global

position information it can in essence

take out that drift error that I was

talking about where the aircraft might

not know exactly where they are in the

world but they will know where they are

relative to each other and their

uncertainty about where they are in the

world will be

smaller by exchanging this relative

range information than it would be if

they didn't exchange the information got

it and so another I think realistic

application would be within cities like

if you're flying between sky skyscrapers

then you're going to lose the signal or

it's going to be distorted somehow is

that is that another place to use it

yeah I think essentially anywhere um

that you have um degradation of GPS or

loss of GPS this sort these sort of

Concepts could be exploited yeah so and

then flying in and out of a warehouse

for example Amazon warehouse or whoever

that that could work I've also wondered

so people talk about having uh like UAV

like well they're not uavs but but taxis

like air taxies if you had vehicles um

like large vehicles that aren't flying

in a swarm but they are all in the same

area would they still be able to do this

yeah I think the key thing is is would

they have

some um information about where they are

relative to each other that could be

from a camera knowing bearing you know

what direction are you from me or it

could be from communication signals uh

that would give you information about

how far away someone is from you okay so

let's talk about the aspect of having

multiple Vehicles here what if there

could be just one vehicle and it could

map on the ground and that would be

beneficial to it right yes but here is

it um uh it's just to improve the

resolution or to reduce errors that that

you're having multiple Vehicles talk to