AJIP and NOC intersections are located in Bahir Dar city which are labelled as intersection locations by which more frequent and severe RTCs occurred of the city. This thesis aims to introduce Agent Based Modelling and real time simulation for traffic flow analysis in order to identify the traffic factors leading to crashes and contributing to crash severity at the intersections so as to provide with alternative strategy to enhance safety in the intersections. Types of observation data are flow and spot speed that are observed during peak hours in the morning and afternoon. Calibration process is done by minimizing root mean square error of speed between observation and calibrated model. SUMO version 1.14.0 is used to perform the microsimulation. After the model is calibrated and validated, alternative strategy to enhance road safety are proposed for each intersection, namely Scenario 1: providing restricted lane for motorcycle and cycle at NOC intersection and Scenario 2: restriction of heavy goods vehicles passing through AJIP intersection at peak hours. For both Scenarios, the results from SUMO before applying the strategy, will be compared with the results after applying the strategy. The results suggest that the risk of collision can be decreased by 15% when providing scenario 1 for NOC intersection while for scenario 2 in AJIP intersection, the simulation data suggested that we can decrease the risk of collision by 25% with providing this strategy. this result gives more confidence in the validity of the model. Keywords: ABM microsimulation, SUMO, unsignalized intersections, root mean square error, min gap, spot speed, safety