Hacker News new | past | comments | ask | show | jobs | submit login
How Cockroaches Crash into Walls and Keep Going (nytimes.com)
65 points by IntronExon on Feb 14, 2018 | hide | past | favorite | 21 comments

"How cockroaches crash into walls and keep going"

Well, being 20mm long and weighing a couple of grams probably helps. And the "robot cockroach" looks almost nothing like a real cockroach, it's just a scaled down version of RHEX.

Also, the video of the cockroach doesn't show it crashing into the wall. It shows it running towards the wall, adjusting its body posture and stride even before its antennae make contact, slowing abruptly (watch the curved antenna bend forward under the deceleration), then stepping up onto the wall. It's categorically not showing what the text describes. Maybe they have other videos?

The original article [1] does include a video of cockroaches running into the wall without even trying to avoid it, which is somewhat funny.

[1] http://rsif.royalsocietypublishing.org/content/15/139/201706...

Ah, thanks for that! The link to the paper in TFA came up as a 403 when I looked yesterday.

This is a much more interesting video. It looks like (1) the cockroaches head/eyes are robust enough to just take the hit, which is interesting - I always wondered how robust insect eyes were compared with mammalian eyes which are obviously very delicate, and (2) the kinematics of the head/thorax flick the thorax upwards on impact. There's a lot of this kind of 'physical intelligence' built into animals' bodies, and it's a fascinating field of study - witness the recent finding that flamingos on one leg are statically stable: https://www.theatlantic.com/science/archive/2017/05/flamingo...

> Also, the video of the cockroach doesn't show it crashing into the wall. It shows it running towards the wall,

Yea that's the "angling their head upward and using their legs to slow down before reaching the wall" roach. They didn't include any footage of the "start climbing with your face" roach.


HN guidelines, in their infinite wisdom, ask:

Please don't insinuate that someone hasn't read an article. "Did you even read the article? It mentions that" can be shortened to "The article mentions that."


His point is fair, though. The article describes two types of approaches taken by the roaches. But only shows the "clever" variation, with the angled body.

The main finding was that even when the roaches did not angle their bodies (like the robot in the video) they still managed to transition and scale the wall at roughly the same speed. Which is interesting, because it means their physical forms are just so well adapted they just aim their bodies and run and hope for the best and it usually works out.

So ultimately, it is weird that they didn't compare the rough-and-tumble cockroach approach to the robot video... it's weird that they used the "clever cockroach" video juxtaposed with the "dumb robot" video. Didn't really make a compelling point, did it?

Agreed, it's such a strange decision unless for some reason they requested but didn't receive the "crash into a wall" cockroach video. It's confusing others too: https://news.ycombinator.com/item?id=16373846

>> It turns out the cautious approach wasn’t necessary. The roaches that ran headlong into the wall could make the upward shift just as quickly — in about 75 milliseconds — the researchers found.

The opposite is also true though- the dumb, headlong-crash approach isn't necessary either, since it's not faster than the cautious approach.

I think what the article is really saying is that, absent a method to make the cautious approach available, then the dumb approach is good enough.

However, the cautious approach might be evidence of a capacity for more complex behaviours when necessary, the absence of which can significantly hinder the animal's (or, indeed, the robot's) versatility in other situations, besides climbing a wall quickly.

In other words, maybe cockroaches are hard to hit not because they can run headlong at a wall without slowing down but because they have the choice of not doing so when needed.

Which in turn means that robots that can only scale a wall by running blindly at it will still not be as good as cockroaches in tasks other than climbing walls.

The head-first strategy may allow for higher speed before the transition. From the journal article:

> Head-first impact is the primary (approx. 80%) transition strategy and often occurs at higher wall-approach speeds

No comments?

I sure don't miss NYC cockroaches. They're seriously tough mother fuckers. Whack them with a shoe sole and they usually shrug it off and barely break stride.

The real problem with apartments in cities is that your neighbors' roaches become your roaches. No way to keep them from moving from apartment to apartment via cracks in the walls. I don't know how you prevent something like that? I'm now in a single family detached house in suburbia. That has its own problems, but roach infestations aren't among them.

I used to live in a connected house in Brooklyn. In fact, the whole neighborhood is streets of ten to twenty all-brick, connected houses. So if even one person had a roach problem, it could migrate a couple houses down. Having cats would definitely help.

One time, our neighbors decided to move and had moving sales. Well, apparently their house was so decrepit and disgusting, the stove just covered in food and soot, roaches roaming freely everywhere. We had seen a roach or two in our house, a waterbug maybe, but were surprised that we hadn't seen more considering how close it all was. I guess the brick/cement walls were tight?

I grew up in that house for 20 years, so I'm partial to having fond memories despite my mother always complaining about how its' too little square feet and too many steps. We were robbed when I was little, I have no recollection except lucid and possibly ~fake~ dreams. My parents responded by putting wrought iron bars on all the windows. Smart move. In fact I was often afraid of being mugged since I had to get on the bus to goto Brooklyn Tech at around 5am when it was dark. But I survived. Used to carry bear-grade pepper spray and a nice sized SpiderCo Civilian defense knife. Never had to use them thankfully.

We had a basement that we finished nicely, so it was 3 decent sized floors - but so many, so many steps, two long flights. My parents still live in that house but want to move. I believe they purchased it nearly 35 years ago for close to $150k. Now it's worth $800k. And its' not a mansion, brownstone, or anything fancy. That's NYC / Brooklyn for you ;-)

I feel like if I carried a knife I'd just be that more likely to cut myself on it, or have it taken and used against me...

Don't carry a weapon if you are not prepared to use it. Very important dogma.

> No way to keep them from moving from apartment to apartment via cracks in the walls.

Not totally effective because it's almost impossible to be thorough enough, but you can use painters caulk to seal up every crevice you can find. You'll also need weatherstrip around and a seal under your exterior doors.

Unlike rodents, roaches don't gnaw through walls.

Regular applications of insecticide help also but a lot of people don't like that.

Neat! I recently attended a talk by Jayaram about his work with cockroaches: it's fascinating stuff, especially on a mechanical level.

As someone else mentioned, size and weight are huge factors in kinetic design. Very roughly: when you're smaller than the dimensions of a mouse, you can crash into things and count on passive mechanisms to recover. The relative impact of crashes increases as you scale up, and so it becomes more worthwhile to have a neurological mechanism that can sense obstacles beforehand and come up with other ways to avoid them. Timing these neural signals also becomes an issue that scales with size. So cockroaches and other small animals can use very different movement-control mechanisms than we do, and the same scaling principle would apply to cockroach- or human-sized robots.

For the cockroach in the first video, it's definitely sensing the wall and preparing itself to change trajectory. But cockroaches who just crash into walls and flail blindly to recover don't actually do worse than those who take the time and energy to anticipate the walls. This is because they're at the right size and material composition so that the crash doesn't debilitate them.

And yes -- there are many, many more videos that Jayaram took as part of his work. Unfortunately I can't find any wall-crashing videos, but there are some squeezing videos here: http://www.pnas.org/content/113/8/E950/tab-figures-data

I'm not a roboticist nor do I hold a PhD, but the robot cockroach wasn't compelling to me at all. The cockroach very clearly elevates its posture as it approaches the wall. The robot just brute-forces contact until its legs manage to catch hold. Really no similarity at least from the videos.

That being said, the study is still very interesting!

They showed only one version of cockroach. Not one which crashes.

According to me, the antennae of the cockroach play a vital role in sensing the obstacle ahead; but the work seems to make no use of this fact. There needs to be some use of the frontal information in making the guided decision, prior to struggling to climb the wall like the robot does.

I'm already scared of Boston Dynamics. Once their bots start to look like cockroaches, I'll never sleep the same.

Meh. Nothing grosses me out as much as an actual cockroach. I consider myself a pretty rational and relatively brave person, but every time I see one way up on a wall I retreat to a different room in the off-chance that it decides to jump and fly straight into me (which has happened before.)

Roaches are pretty harmless other than being disgusting looking.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact