Distributed computer system and method enabling application of autonomous agents
View Patent ↗Disclosed is distributed computer system enabling application of autonomous agents (AAs) across domains, comprising: primary distributed ledger arrangement to implement software framework that comprises: client-agent device (client-AA) to receive service request (SR), generate objective associated with SR, send objective to agent device (AA); processing arrangement to create clusters of AAs based on similarity of their functionalities; secondary distributed ledger arrangement (SDLA) to coordinate collective learning of AAs by sharing model updates in clusters, AAs in cluster is configured to train computing(s) model, SDLA comprises: first secondary distributed ledger (FSDL), comprising language model module to send context of SR, second secondary distributed ledger (SSDL), comprising machine learning model agent (MLMA) to: send list and protocol specification (PS) update to AA. AA comprises context builder software module to obtain list, PS module to generate PS for fulfilment of SR, and build executor module to compose AAs associated into further autonomous agent (further-AA), further-AA implement PSs to fulfil SR by client-AA.
1 . A distributed computer system that enables application of autonomous agents across a plurality of domains, the distributed computer system comprising a plurality of modular and extensible software modules configured to operate as a plurality of autonomous agents comprised within a plurality of computing devices, wherein the autonomous agents are communicably coupled to each other, the distributed computer system further comprising:
a primary distributed ledger arrangement that implements a software framework, the software framework comprising a client-agent device and an agent device, wherein the client-agent device is configured to receive a service request, generate an objective associated with the service request, and provide the objective to the agent device;
a processing arrangement configured to form and maintain clusters of the plurality of autonomous agents according to similarity metrics applied to stored agent data and/or stored agent models that define agent capabilities and task-handling characteristics;
a secondary distributed ledger arrangement configured to store and update collective learning state information for the plurality of autonomous agents, wherein the secondary distributed ledger arrangement is configured to store model updates generated by autonomous agents within a cluster independently of underlying training data used to generate the model updates, and wherein the secondary distributed ledger arrangement is configured to make the stored model updates selectively available to autonomous agents belonging to the same cluster;
wherein the secondary distributed ledger arrangement comprises:
at least one first secondary distributed ledger comprising a language model module configured to generate contextual information associated with the service request communicated via the agent device; and
at least one second secondary distributed ledger comprising a machine-learning model agent configured to identify, in accordance with the objective, a set of autonomous agents associated with the objective and to provide a protocol-specification update associated with the objective;
wherein the agent device comprises:
a context-builder software module configured for data exchange with the machine-learning model agent to obtain the identified set of autonomous agents;
a protocol-specification module configured to generate at least one executable protocol in accordance with the protocol-specification update; and
a build-executor module configured to compose the autonomous agents of the identified set into a further autonomous agent,
wherein the further autonomous agent is configured to execute the at least one executable protocol to fulfil the service request.
2 . The distributed computer system of claim 1 , wherein the processing arrangement is configured to form and maintain the clusters by applying similarity metrics to stored agent data and/or stored agent models representing at least one of historical service requests handled by the autonomous agents, domain identifiers associated with the autonomous agents, or parameters of agent-specific computing models.
3 . The distributed computer system of claim 2 , wherein the processing arrangement is configured to apply a loss matrix representing distances between agent data and/or agent models to form and maintain the clusters.
4 . The distributed computer system of claim 1 , wherein the secondary distributed ledger arrangement is configured to store model updates generated by autonomous agents independently of underlying training data used to generate the model updates, and to exclude storage of the underlying training data on the secondary distributed ledger arrangement.
5 . The distributed computer system of claim 1 , wherein the protocol-specification update defines at least one of message formats, execution sequencing rules, or acknowledgment requirements governing interactions between the autonomous agents, and wherein the protocol-specification module is configured to generate the at least one executable protocol as a machine-interpretable data structure enforceable by the further autonomous agent.
6 . The distributed computer system of claim 1 , wherein the build-executor module is configured to compose the further autonomous agent by establishing cryptographic associations between the autonomous agents of the identified set, and wherein the further autonomous agent is configured to enforce access control to the at least one executable protocol using encryption.
7 . A method for enabling application of autonomous agents across a plurality of domains, the method comprising:
receiving, at a client-agent device of a software framework implemented by a primary distributed ledger arrangement, a service request;
generating, by the client-agent device, an objective associated with the service request;
providing the objective from the client-agent device to an agent device;
forming and maintaining, by a processing arrangement, clusters of a plurality of autonomous agents according to similarity metrics applied to stored agent data and/or stored agent models that define agent capabilities and task-handling characteristics;
storing and updating, by a secondary distributed ledger arrangement, collective learning state information for the plurality of autonomous agents, including storing model updates generated by autonomous agents independently of underlying training data used to generate the model updates and making the stored model updates selectively available to autonomous agents belonging to the same cluster;
generating, by at least one first secondary distributed ledger of the secondary distributed ledger arrangement using a language model module, contextual information associated with the service request communicated via the agent device;
identifying, by at least one second secondary distributed ledger of the secondary distributed ledger arrangement using a machine-learning model agent and in accordance with the objective, a set of autonomous agents associated with the objective, and providing a protocol-specification update associated with the objective;
obtaining, by a context-builder software module of the agent device, the identified set of autonomous agents via data exchange with the machine-learning model agent;
generating, by a protocol-specification module of the agent device, at least one executable protocol in accordance with the protocol-specification update;
composing, by a build-executor module of the agent device, the autonomous agents of the identified set into a further autonomous agent; and
executing, by the further autonomous agent, the at least one executable protocol to fulfil the service request.
8 . The method of claim 7 , wherein forming and maintaining the clusters comprises applying similarity metrics to stored agent data and/or stored agent models representing at least one of historical service requests handled by the autonomous agents, domain identifiers associated with the autonomous agents, or parameters of agent-specific computing models.
9 . The method of claim 8 , wherein forming and maintaining the clusters comprises applying a loss matrix representing distances between agent data and/or agent models.
10 . The method of claim 7 , wherein maintaining collective learning state information comprises storing model updates generated by autonomous agents independently of underlying training data used to generate the model updates, and excluding storage of the underlying training data on the secondary distributed ledger arrangement.
11 . The method of claim 7 , wherein generating the at least one executable protocol comprises generating a machine-interpretable data structure that defines at least one of message formats, execution sequencing rules, or acknowledgment requirements governing interactions between the autonomous agents, the machine-interpretable data structure being enforceable by the further autonomous agent.
12 . The method of claim 7 , wherein composing the further autonomous agent comprises establishing cryptographic associations between the autonomous agents of the identified set, and wherein executing the at least one executable protocol comprises enforcing access control using encryption.