Insect navigation and path finding Tobias Seidl∗ January 18, 2008 1 Abstract design. Insects have a brain weighing about a tenth of a milligram. Nevertheless some in- 2 Introduction sect species exhibit amazing performance in finding their daily paths when looking Evolutionarily challenged organisms opti- for food or shelter. Some species have even mize among others in terms of energy ex- found their ecological niche in being excel- penditure (e.g. avoiding detours), informa- lent navigators and hence can survive in ex- tion processing (e.g. energy, reliability), treme habitats. One remarkable and well safety (e.g. risk avoidance), failure avoid- studied model is the Saharan desert ant ance (e.g. simplicity, redundancy, failure Cataglyphis fortis that outruns its competi- tolerance). While the qualitative perfor- tors by performing egocentric navigation. mance is similar - the approach is differ- By using skylight cues as a compass and ent. It will be my mission to introduce you counting steps it is able to find its way to the study field of neuroethology (behav- without using external visual cues. In com- ior associated with sensory information ), parison, the Australian desert ant Melopho- to outline the lessons learnt on insect navi- rus bagoti employs route learning strate- gation so far and finally present some rough gies, where it visually learns and in tests re- ideas for space related technical application. calls every point of their route. In studying the insects’ strategies we can learn a great 2.1 Insect cognition deal on how little information can be used to perform a navigational task. Also the Animal orientation can be differentiated way of how information is processed within into two big fields: While migration is deal- such a tiny brain is intriguing. The conclu- ing with the relocation of an animal - some- sions drawn from this research are nowa- times over many generations - from one days not only used to understand human place to another, (ii) navigation denotes the behavior but find their way into technical systematic return to a previously left point ∗Advanced Concepts Team, ESA/ESTEC, Ke- of reference. Both types are found in differ- plerlaan 1, 2201 AN Noordwijk, The Netherlands, ent taxa, ranging from bacteria to whales. Email: [email protected] In the following I will focus on the cen- 1 tral place navigation techniques as found 2.2 Insects qualify in several in insects for a couple of reasons. The in- aspects as models sect model is a very variable one and many different ”technical solutions” are realized When reviewing the panel of this workshop by them. The system architecture of in- you will find two two other contributions sects is a rather simple one compared to on insects as model animals for technically vertebrate models and in consequence eas- oriented studies and application. In a way, ier to understand: First of all the hard- this stresses the fact of the amazing suc- ware (i.e. brain) is very small and hence cess of the insect Bauplan both for the vari- simple. Secondly, the behavioral patterns ability of the exoskeleton and the perfor- are reduced to the elementary motivations: mance neuronal system. While mechani- defense (fight), maintenance (food intake), cal aspects will be dealt by Stanislav Gorb, and reproduction. Since we deal with so- Nicolas Franceschini will introduce the in- cial insects, where the queen suppresses the sects’ visual autopilot. Using both mechan- workers’ reproductive behavior the discrim- ical and neuronal tools, the insects become ination of the two remaining motivations is able to perform high level tasks - also called much simpler than when studying a verte- insect intelligence or insect cognition - such brate model with its huge variety of moti- as navigation. In combining the results of vations and behaviors. In many cases, the these complementary approaches, we will food items found during a foraging trip can- not only learn more about the biological not be eaten by the insect itself. In conse- model, but also may be able to transfer this quence the mission goal becomes a rather knowledge to technical design. abstract one, where the agent-ant has to decide if the item found fulfills the profile set by the super-organism. On top of that, 3 Mastering egocentric it is technically easy to work with insects and so insect models have become very pop- navigation: the Saha- ular among behavioral scientists. Central ran ant Cataglyphis place foraging on the other hand is the type of behavior that involves highly elaborated One of the best studied models for in- signal processing, both in terms of preci- sect navigation is the Saharan desert ant sion and time, and hence appears to be the Cataglyphis. Several Cataglyphis species technically more interesting system. While have found their ecological niche in areas a migratory animal will interrupt its behav- that seem devoid of live: Desert areas con- ior once a suited living ground is found, a taining almost no vegetation and even less navigating forager is forced to locate the ex- animal live seem not to provide accept- act point of reference - usually the nest - in able living conditions. Nevertheless, once order to fulfill its mission. spending some time, the observer will dis- cover numerous rather large (approx 1 cm body length) ants with quite a distinct body shape and a rather fascinating behavior: In 2 this featureless habitat, they run at rather ondly - evaporate the pheromones too fast high speeds (up to 1 m/s) in a seemingly in order to allow for a stable marking. And determined manner toward an invisible goal - thirdly - there are almost not abundant at the end of their trajectory. Indeed, the food sources available that would require ant will eventually disappear in a small in- or justify establishing a route. Food items conspicuous hole in the ground - the en- are sparsely distributed and hence can usu- trance to the colony’s nest, hosting up to a ally be carried away by a single ant. Pass- few thousand individuals. A deeper study ing on the knowledge of a feeding site to will reveal, that emerging from this nest, a nest mate would not contribute to the a cohort of ants will go foraging for food colony’s success. On the other hand, each throughout the day, performing one of the ant that can be found outside the nest fol- most fascinating behaviors in animal king- lows the very same programme of search- dom. Every little ant will leave the nest for ing for food and then safely and quickly a distance of up to twenty thousand times navigating home. Hence desert ants are their own body length, and search for a sin- an ideal study model for (biological) au- gle food item - a dead insect for example tonomous agents. that succumbed to the atrocious climatic conditions. Once successful, the ant will 3.1 The compass turn around and in a determined manner steer back home to the nest in a direct line Cataglyphis ants navigate egocentrically, without getting lost. Even in the absence i.e. without using external visual cues. This of any visual cue, i.e. landmarks, the ants implies that they constantly monitor their will successfully locate their nest and imme- own movements and update the knowledge diately initiate another foraging run. This of their position in respect to the nest, the behavior is somewhat different to the com- home vector. Path integration - as the ba- monly known routes that can be observed in sis of vector navigation - requires knowl- the majority on the European ant-species. edge of the direction, distance and incli- Wood ants for example establish a route to nation of the current segment of the path a feeding site, i.e. a dead mouse, by placing in order to calculate the home-vector. The olfactory marks along the way. Other ants by far best studied element of desert ant will follow this route and easily find the food navigation is the compass mechanism with source. Each of the ants will add its olfac- which the ants determine rotation around tory share to the route and in consequence, the vertical axis. The work on the sky- the route will be reinforced and even attract light compass of insects started with von more ants, until the source is getting ex- Frisch’s [3] Nobel-awarded discovery that ploited. Now this approach does not work bees use polarized light to determine flight with desert ants for several reasons. Firstly, directions. In short, many insects, includ- the light sand in the desert has little cohe- ing Cataglyphis have a part of their com- sion and hence the route marked onto it will pound eye specifically adapted for compass be carried away easily by wind. If not the tasks. The dorsal rim region, i.e. a small wind, the dry and hot climate will - sec- part on the very top of the head, has eyes 3 that are sensitive to polarized UV-light of ing the energy spent, or the time spent to a preferred angle arranged in a fan shape cover a certain distance. Both were dis- configuration [15, 2, 12]. Theoretical con- carded for various reasons and hence only siderations propose the existence of three exploiting visual cues and proprioceptive integrator neurons that are tuned to dif- ones (i.e. own movement) remained. While ferent directions and subsequently pass on flying insects such as bees use optic flow for the information of direction to so called optometry, this factor plays only a minor (and proposed) compass neurons. However, role in running animals (16%, [8, 9]). A there are some apparent constraints when series of studies finally demonstrated, that exploiting the polarization of the sky for distance estimation is performed by a pe- directional purposes.