Application of learning pallets for real-time scheduling by the use of radial basis function network

The expansion of the scope and scale of products in the current business environments causes a continuous increase in complexity of logistics activities. In order to deal with this challenge in planning and control of logistics activities, several solutions have been introduced. One of the most latest one is the application of autonomy. The paradigm of autonomy in inbound logistics, can be reflected in decisions for real-time scheduling and control of material flows. Integration of autonomous control with material carrier objects can realize the expected advantages of this alternative into shop-floors. Since pallets (bins, fixtures, etc.) are some common used carrier objects in logistics, they have the potential to undertake the responsibility of real-time jobs dispatching to machines in the shop-floor scheduling problem. Hence, the current paper covers the problem of real-time scheduling in a stochastic and complex shop-floor environment, by means of autonomy. Indeed, the sustainment's advantage of pallets in manufacturing systems has inspired the idea of developing learning pallets (Lpallets) with the capability of autonomous control in complex and uncertain logistics environment with abrupt changes. Among some intelligent techniques, the artificial neural network (ANN) and, specially, the radial basis function network (RBFN) is selected to transmit the abilities of intelligent decision-making as well as learning to Lpallets in a distributed manner. Some variants in training and RBFN application alternatives are considered to evaluate the competency of RBFN for Lpallets. An Lpallet makes its dispatching and control decision based on its own experience and intelligence about the entire system and local situations in an exemplary hybrid flow-open shop problem. To prove the claimed application of RBFN in autonomous Lpallets a discrete-event simulation model is developed for the assembly scenario.