Periodic Reporting for period 3 - ANTSolve (A multi-scale perspective into collective problem solving in ants)
Período documentado: 2021-06-01 hasta 2022-11-30
It is well known that group living holds many advantages but, is collective intelligence one of these? Is a group really more intelligent than its smartest individual? This question is not well-defined and thus difficult to address. In this project we approach such questions by studying collective problem solving by groups of ants.
Ants coordinate their action in a distributed way without strong hierarchies nor centralized decision makers. In today’s world, society is heading in similar directions: the internet, social media, and even electric networks are becoming more and more decentralized. However, as recent history teaches us functioning in a collectively advantageous manners within this setting is far from trivial. It is therefore highly interesting and relevant to study how the ants who have 150 million years of experience on this issues cooperate to collectively face large scale problems.
Our overall objective is to formulate a multi-scale description of problem solving by Paratrechina longicornis crazy ants. The problems we confront the ants with are natural – cooperatively transporting large objects to their nest. We challenge the ants with a variety of obstacles and confinements and measure how they react to them as individuals and as groups. We then link these reactions to the concept of problem solving wherein one can define group success rates and efficiency. Further, by duplicating some of the ant mazes on a much larger scale we aim to compare the problem solving capabilities of ants groups to those of human groups. Finally, inspired by these results we will use modelling and theory to broaden our definitions and understandings of the concept of collective cognition.
Ants coordinate their action in a distributed way without strong hierarchies nor centralized decision makers. In today’s world, society is heading in similar directions: the internet, social media, and even electric networks are becoming more and more decentralized. However, as recent history teaches us functioning in a collectively advantageous manners within this setting is far from trivial. It is therefore highly interesting and relevant to study how the ants who have 150 million years of experience on this issues cooperate to collectively face large scale problems.
Our overall objective is to formulate a multi-scale description of problem solving by Paratrechina longicornis crazy ants. The problems we confront the ants with are natural – cooperatively transporting large objects to their nest. We challenge the ants with a variety of obstacles and confinements and measure how they react to them as individuals and as groups. We then link these reactions to the concept of problem solving wherein one can define group success rates and efficiency. Further, by duplicating some of the ant mazes on a much larger scale we aim to compare the problem solving capabilities of ants groups to those of human groups. Finally, inspired by these results we will use modelling and theory to broaden our definitions and understandings of the concept of collective cognition.
We have made progress on several fronts.
On the infrastructure side, we have constructed two outdoor facilities on Weizmann Institute grounds. One is the ANT platform that allows us to survey the cooperative transport process over large areas using multiple cameras. The second is the human-maze area where we film people as they maneuver through tight sport either alone or in a group.
On the single ant scale we have consolidated the fact that individual carriers continuously react to the group rather than simply pull towards the nest. These results have been published in the journal Nature Physics. Further, we have tracked individual ants during the cooperative transport process. We tracked the behavior of ants before and after they reach a food source. After reaching the food the ants may transport it on its own (if it is small enough), transport it as a part of a group, recruit more ants to it while laying scent marks, or clear out obstacles in the vicinity of the moving load (a new phenomenon we now discovered). We correlate their actions and motion trajectories in these two cases. We are now analyzing these trajectories to understand how the terrain and the ants knowledge about it affect her internal representations of it and the therefore her future actions. We further designed a force measuring apparatus that can measure the tiny forces exerted by individual ans. We use this apparatus to test our microscopic model for ant behavior and to gain new insights into how individual ants join into the collective hauling effort.
On the multiple ant scale we have now developed and constructed a cooperative transport robot, we used this robot to experimentally simulate the reaction of the carrying group to different mechanical stimuli (e.g. hitting an immobile obstacle, or the joining of a new ant). This data is currently under analysis. Further, we have collected large amounts of data of the ants carrying a load between the same two points. Using this data we are constructing intermediate scale, differential equation based methods for describing the ants cooperative transport movement. Last, we have filmed the cooperative transport process over large areas where we can track the history and actions of all ants throughout the transport process.
In terms of problem solving – we are advancing in three fronts. One is having the ants travel through semi-natural terrains while carrying a large objects. These terrains were created by dispersing cubes at different densities in the ants’ path and allowing them to move through these disordered terrains to the nest. We measured the ants traversal times at different cube densities and by using an analogy to the mathematical/physical theory of Percolation were able to show that the ants are highly efficient. We have further shown that they achieve this efficiency using minimal resources. These results were published in the journal eLife. Further, we studied the behavior of different group sizes as they face a binary choice – crossing an obstacle either from the left or from the right. We found that the decision making process, and specifically the weight given to the minority opinion, is highly dependent on group size. We used a modelling approach to connect the ants’ collective behavior to decision making in the brain and suggest extensions to the classical neuroscientific models that would render them much more general. A manuscript describing this work has been submitted. Finally, we have followed both ants and people as they navigate large objects through tight sport at different group size. This allows us to compare problem solving capabilities across species and group sizes.
On the modelling front, we are now uniting our existing agent based models with a physics engine. This will allow us to create predictions for the ants’ motions through complex terrains (as required for problem solving) and use these two find the correct ways to extend our model.
On the infrastructure side, we have constructed two outdoor facilities on Weizmann Institute grounds. One is the ANT platform that allows us to survey the cooperative transport process over large areas using multiple cameras. The second is the human-maze area where we film people as they maneuver through tight sport either alone or in a group.
On the single ant scale we have consolidated the fact that individual carriers continuously react to the group rather than simply pull towards the nest. These results have been published in the journal Nature Physics. Further, we have tracked individual ants during the cooperative transport process. We tracked the behavior of ants before and after they reach a food source. After reaching the food the ants may transport it on its own (if it is small enough), transport it as a part of a group, recruit more ants to it while laying scent marks, or clear out obstacles in the vicinity of the moving load (a new phenomenon we now discovered). We correlate their actions and motion trajectories in these two cases. We are now analyzing these trajectories to understand how the terrain and the ants knowledge about it affect her internal representations of it and the therefore her future actions. We further designed a force measuring apparatus that can measure the tiny forces exerted by individual ans. We use this apparatus to test our microscopic model for ant behavior and to gain new insights into how individual ants join into the collective hauling effort.
On the multiple ant scale we have now developed and constructed a cooperative transport robot, we used this robot to experimentally simulate the reaction of the carrying group to different mechanical stimuli (e.g. hitting an immobile obstacle, or the joining of a new ant). This data is currently under analysis. Further, we have collected large amounts of data of the ants carrying a load between the same two points. Using this data we are constructing intermediate scale, differential equation based methods for describing the ants cooperative transport movement. Last, we have filmed the cooperative transport process over large areas where we can track the history and actions of all ants throughout the transport process.
In terms of problem solving – we are advancing in three fronts. One is having the ants travel through semi-natural terrains while carrying a large objects. These terrains were created by dispersing cubes at different densities in the ants’ path and allowing them to move through these disordered terrains to the nest. We measured the ants traversal times at different cube densities and by using an analogy to the mathematical/physical theory of Percolation were able to show that the ants are highly efficient. We have further shown that they achieve this efficiency using minimal resources. These results were published in the journal eLife. Further, we studied the behavior of different group sizes as they face a binary choice – crossing an obstacle either from the left or from the right. We found that the decision making process, and specifically the weight given to the minority opinion, is highly dependent on group size. We used a modelling approach to connect the ants’ collective behavior to decision making in the brain and suggest extensions to the classical neuroscientific models that would render them much more general. A manuscript describing this work has been submitted. Finally, we have followed both ants and people as they navigate large objects through tight sport at different group size. This allows us to compare problem solving capabilities across species and group sizes.
On the modelling front, we are now uniting our existing agent based models with a physics engine. This will allow us to create predictions for the ants’ motions through complex terrains (as required for problem solving) and use these two find the correct ways to extend our model.
Progress beyond state of the art was achieved on several fronts:
• Collecting complete data regarding cooperative transport over large areas (several meters square).
• We have constructed a cusumized load that allows for the measurement of single ant forces down to the milli-Newton scale.
• We have constructed an ant robot that allows for precise mechanical interactions with the carrying ants.
• We have discovered a new biological phenomenon wherein the ants clear beads from the path of the moving object.
• We have constructed a new setup that allows us to challenge ant groups of different sizes from one up with the precise same clearly defined geometrical riddles. This was never done before.
• We have confronted human groups of different sizes with the same problems.
• We have tracked single ant navigation and pheromone trailing in a controlled information setup.
• We tracked ant groups of different sizes as they confronted binary decisions. We show a dynamic decision making process and use it to expand current models for decision making suggested in the field of neuroscience.
• For the first time ever, we show how real ants tackle the famous ant-in-a-labyrinth problem. We demonstrate how the ants surpass known physics inspired solutions to this problem.
• Collecting complete data regarding cooperative transport over large areas (several meters square).
• We have constructed a cusumized load that allows for the measurement of single ant forces down to the milli-Newton scale.
• We have constructed an ant robot that allows for precise mechanical interactions with the carrying ants.
• We have discovered a new biological phenomenon wherein the ants clear beads from the path of the moving object.
• We have constructed a new setup that allows us to challenge ant groups of different sizes from one up with the precise same clearly defined geometrical riddles. This was never done before.
• We have confronted human groups of different sizes with the same problems.
• We have tracked single ant navigation and pheromone trailing in a controlled information setup.
• We tracked ant groups of different sizes as they confronted binary decisions. We show a dynamic decision making process and use it to expand current models for decision making suggested in the field of neuroscience.
• For the first time ever, we show how real ants tackle the famous ant-in-a-labyrinth problem. We demonstrate how the ants surpass known physics inspired solutions to this problem.